Directly access YottaDB data from C #59
Comments
estess
added a commit
to estess/YottaDB
that referenced
this issue
Nov 9, 2017
…le API - part 1
No major functionality add with this drop. The primary addition to the changes
mentioned in the README is that error ids have been changed from global ints
to simple #define values so they are constants instead of global variables.
Also retrofitted all changes to the new assembler modules for the newly supported
ARM platform.
Lastly, a regression or two with call-ins have been fixed and a few edge-case
day-one issues also with call-ins have been fixed:
1. HALT no longer halts but returns to the caller
2. ZHALT no longer halts and returns its argument to the caller.
3. QUIT with no arg when a return value was requested raises an error.
4. QUIT with an arg when no return value was expected raises an error.
5. Call with args to a routine that has none defined raises an error.
6. Various fixes dealing with return values when HALT/ZHALT/ZGOTO 0 are
unwinding the stack that exploded were fixed.
Changes per module:
- All sr_armv7l/ assembler modules
- Had cosmetic changes like putting all assembler directives at the first tab
instead of in column zero to make things more standardized.
- All sr_armv7l/ AND sr_x86_64/ assembler modules
- The .include for debug.si is changed to a #include so the C preprocessor pulls
it in instead. This allows the ERR_GTMCHECK reference in the macro definitions
in this include to get their numeric substitutions from "ydbmerrors.h".
- sr_armv7l/ci_restart.s
- removed as no longer needed
- sr_armv7l/cmerrors_ctl.c
- sr_x86_64/cmerrors_ctl.c
- Minor message change (in .msg file) to force recreation of this _ctl file.
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/cmierrors_ctl.c
- sr_x86_64/cmierrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/debug.si
- sr_x86_64/debug.si
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/dm_start.s
- sr_x86_64/dm_start.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/gdeerrors_ctl.c
- sr_x86_64/gdeerrors_ctl.c
- Minor message change (in .msg file) to force recreation of this _ctl file.
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/libydberrors.h
- sr_x86_64/libydberrors.h
- New file with #defines of error ids prefixed by YDB_ and with negative values that
are used in the new simple API.
- sr_armv7l/merrors_ctl.c
- sr_x86_64/merrors_ctl.c
- Minor message changes (in .msg file) to force recreation of this _ctl file.
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/op_exfun.s
- sr_x86_64/op_exfun.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_extcall.s
- sr_x86_64/op_extcall.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_extexfun.s
- sr_x86_64/op_extexfun.s
- Change ERR_GTMCHECK & ERR_FMLLSTMISSING access from global int to #define value
- sr_armv7l/op_extjmp.s
- sr_x86_64/op_extjmp.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofexfun.s
- sr_x86_64/op_mprofexfun.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofextcall.s
- sr_x86_64/op_mprofextcall.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofextexfun.s
- sr_x86_64/op_mprofextexfun.s
- Change ERR_GTMCHECK & ERR_FMLLSTMISSING access from global int to #define value
- sr_armv7l/ydbcmerrors.h
- sr_x86_64/ydbcmerrors.h
- Remove stuff that belongs in cmerrors_ctl.c (invalid generation and botched rebase).
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydbcmierrors.h
- sr_x86_64/ydbcmierrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydberrors.h
- sr_x86_64/ydberrors.h
- Remove stuff that belongs in ydberrors_ctl.c (invalid generation and botched rebase).
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydberrors_ctl.c
- sr_x86_64/ydberrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/ydbgdeerrors.h
- sr_x86_64/ydbgdeerrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydbmerrors.h
- sr_x86_64/ydbmerrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_linux/platform.cmake
- Add specific include option for ydbmerrors.h for all assembler routines in both
armv7l and x86_64 platforms.
- Add option so all assembler routines are first processed with the C pre-processor
to pick up error id definitions and do appropriate constant substitution.
- Change the names of the entry points that we link with since all call-in entry point
names changed from gtm_ to ydb_
- sr_port/cmerrors.msg
- Minor message change to force recreation of the associated _ctl file.
- sr_port/cmidef.h
- Add include to pick up new #define style cmierrors definitions.
- sr_port/cmmdef.h
- Add include to pick up new #define style cmerrors definitions.
- sr_port/gdeerrors.msg
- Minor message change to force recreation of the associated _ctl file.
- sr_port/goframes.c
- Change some "spaces" white space to tabs.
- Change the logic where a frame expecting a return code is unwound and a value placed
such that if in a call-in and ZHALT has specified a return value, use that value
instead of the default NULL/0 value. Allows ZHALT to return its value to call-in user.
- sr_port/gtm_common_defs.h
- Remove #defines from no-longer-supported platforms.
- Nullify output of error_def() macro as it is not needed anymore as all error #defines
are always available to all routines that include mdef.h.
- sr_port/gtm_threadgbl_defs.h
- Add gtmci_retval that holds the mval address where a return value is to be set if
the frame is unwound via goframes(). Presently used to return value of ZHALT in
call-in mode.
- sr_port/mdef.h
- Add includes for ydbmerrors.h and ydberrors.h so all error codes are available to
every compilation.
- sr_port/merrors.msg
- Minor message changes to force recreation of the associated _ctl file.
- sr_port/msg.m
- Add an $ETRAP so errors aren't invisible.
- Remove creation of LITDEF entries for each error id in *_ctl.c files.
- Create ydb*.h header files that contain a #define for each error id.
- For merrors.h, also create libydberrors.h that contain a YDB_ prefixed version of the
error name (e.g. YDB_ERR_ACK) and whose value is the negated value of the actual id for
use in the libyottadb API.
- Remove VMS paths.
- sr_port/mumps.hlp
- Fix message output in a ZMESSAGE example.
- sr_port/op_halt.c
- The return code back to dm_start should be SUCCESS (1) instead of zero for a
missing return code to set correctly.
- sr_port/op_zg1.c
- Set the return code to dm_start to be SUCCESS overriding any previous value set.
- sr_port/op_zhalt.c
- Set the return code to dm_start to be SUCCESS instead of the argument value. The
argument value is saved in the gtmci_retval thread global for goframes() to find
and use.
- Return SUCCESS to dm_start().
- sr_port/ydberrors.msg
- Minor message changes to force recreation of the associated _ctl file.
- sr_unix/cmidefsp.h
- Remove #defines for CMI_REASON_{IODONE, INTMSG, CONNECT} since they duplicate definitions
in ydbcmierrors.h and have different values (don't know why).
- sr_unix/errorsp.h
- Mostly cosmetic fix but change include to use quotes instead of angle brackets on name.
- sr_unix/exttab_parse.c
- Relocate static tables from mid-code to top of routine per code standards.
- Rename all gtm_* types to ydb_* types.
- sr_unix/fgncalsp.h
- Rename all gtm_* types to ydb_* types.
- sr_unix/gtm.c
- Remove conditional code for unsupported platforms.
- Remove integer definitions for some errors as these errors are all defined by the
the include pulled in by mdef.h.
- sr_unix/gtm_env.csh
- Change gtm_ci entry point name to ydb_ci in the linker options definition.
- sr_unix/gtm_trigger.c
- Factor out mumps_status initialization this is not a valid value for mumps_status
but it serves to clear out any other (error) value that might be there as we
gear up to start the trigger.
- Added comment to mumps_status initialization to 0. This value does not last as it
is almost immediately reset to SUCCESS (1) by dm_start.
-sr_unix/gtmci.c
- Change static declaration of get_entry to a STATICFNDCL/STATICDEF definition instead.
- Rename all gtm_* types and routines to ydb_* types.
- Initialize gtmci_retval on entry to ydb_ci[p.j]().
- Verify if calling a routine with parms that it has a formal list and raise the
FMLLSTMISSING if not.
- Rearrange some of the case statemments to stack entries doing same action.
- Add call to op_exfunret() to validate/verify the return value if have one.
This raises appropriate errors if the return value is not expected or is otherwise
appropriate.
- sr_unix/gtmci.h
- Rename gtm_ci_exec() to ydb_ci_exec()
- sr_unix/gtmci_signals.c
- Code standard cosmetic changes to put opening brace on its own line.
- sr_unix/gtmexe_symbols.exp
- sr_unix/gtmshr_symbols.exp
- Change call-in related exported names from gtm_* to ydb_*
- sr_unix/gtmxc_types.h
- sr_unix/ydbxc_types.h
- Separated gtmxc_types.h into two parts:
1. ydbxc_types.h - contains the ydb_* call-in routine definitions plus the ydb_* type
definitions.
2. gtmxc_types.h - includes ydbxc_types.h and also includes both the gtm_* types
(typedef'd to ydb types) and the long deprecated xc_* types. In addition it contains
some #defines to turn gtm_* call-in routine invocations into the new ydb_* routines.
- sr_unix/op_fnfgncal.c
- Rename all gtm_* types to ydb_* types.
- sr_unix_gnp/cmierrors.msg
- Minor message changes to force recreation of the associated _ctl file.
- sr_x86_64/gdeerrors_ctl.c
- Minor message change (in .msg file) to force recreation of this _ctl file.
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_x86_64/libydberrors.h
estess
added a commit
to estess/YottaDB
that referenced
this issue
Nov 13, 2017
…commit to libyottadb.h which will hold the needed definitions for the libyottadb simple API as well as the existing call-in, call-out, plug-in etc uses of the types and routines it contains.
estess
added a commit
to estess/YottaDB
that referenced
this issue
Nov 20, 2017
…le API - part 1
No major functionality add with this drop. The primary addition to the changes
mentioned in the README is that error ids have been changed from global ints
to simple #define values so they are constants instead of global variables.
Also retrofitted all changes to the new assembler modules for the newly supported
ARM platform.
Lastly, a regression or two with call-ins have been fixed and a few edge-case
day-one issues also with call-ins have been fixed:
1. HALT no longer halts but returns to the caller
2. ZHALT no longer halts and returns its argument to the caller.
3. QUIT with no arg when a return value was requested raises an error.
4. QUIT with an arg when no return value was expected raises an error.
5. Call with args to a routine that has none defined raises an error.
6. Various fixes dealing with return values when HALT/ZHALT/ZGOTO 0 are
unwinding the stack that exploded were fixed.
Changes per module:
- All sr_armv7l/ assembler modules
- Had cosmetic changes like putting all assembler directives at the first tab
instead of in column zero to make things more standardized.
- All sr_armv7l/ AND sr_x86_64/ assembler modules
- The .include for debug.si is changed to a #include so the C preprocessor pulls
it in instead. This allows the ERR_GTMCHECK reference in the macro definitions
in this include to get their numeric substitutions from "ydbmerrors.h".
- sr_armv7l/ci_restart.s
- removed as no longer needed
- sr_armv7l/cmerrors_ctl.c
- sr_x86_64/cmerrors_ctl.c
- Minor message change (in .msg file) to force recreation of this _ctl file.
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/cmierrors_ctl.c
- sr_x86_64/cmierrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/debug.si
- sr_x86_64/debug.si
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/dm_start.s
- sr_x86_64/dm_start.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/gdeerrors_ctl.c
- sr_x86_64/gdeerrors_ctl.c
- Minor message change (in .msg file) to force recreation of this _ctl file.
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/libydberrors.h
- sr_x86_64/libydberrors.h
- New file with #defines of error ids prefixed by YDB_ and with negative values that
are used in the new simple API.
- sr_armv7l/merrors_ctl.c
- sr_x86_64/merrors_ctl.c
- Minor message changes (in .msg file) to force recreation of this _ctl file.
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/op_exfun.s
- sr_x86_64/op_exfun.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_extcall.s
- sr_x86_64/op_extcall.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_extexfun.s
- sr_x86_64/op_extexfun.s
- Change ERR_GTMCHECK & ERR_FMLLSTMISSING access from global int to #define value
- sr_armv7l/op_extjmp.s
- sr_x86_64/op_extjmp.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofexfun.s
- sr_x86_64/op_mprofexfun.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofextcall.s
- sr_x86_64/op_mprofextcall.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofextexfun.s
- sr_x86_64/op_mprofextexfun.s
- Change ERR_GTMCHECK & ERR_FMLLSTMISSING access from global int to #define value
- sr_armv7l/ydbcmerrors.h
- sr_x86_64/ydbcmerrors.h
- Remove stuff that belongs in cmerrors_ctl.c (invalid generation and botched rebase).
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydbcmierrors.h
- sr_x86_64/ydbcmierrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydberrors.h
- sr_x86_64/ydberrors.h
- Remove stuff that belongs in ydberrors_ctl.c (invalid generation and botched rebase).
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydberrors_ctl.c
- sr_x86_64/ydberrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/ydbgdeerrors.h
- sr_x86_64/ydbgdeerrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydbmerrors.h
- sr_x86_64/ydbmerrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_linux/platform.cmake
- Add specific include option for ydbmerrors.h for all assembler routines in both
armv7l and x86_64 platforms.
- Add option so all assembler routines are first processed with the C pre-processor
to pick up error id definitions and do appropriate constant substitution.
- Change the names of the entry points that we link with since all call-in entry point
names changed from gtm_ to ydb_
- sr_port/cmerrors.msg
- Minor message change to force recreation of the associated _ctl file.
- sr_port/cmidef.h
- Add include to pick up new #define style cmierrors definitions.
- sr_port/cmmdef.h
- Add include to pick up new #define style cmerrors definitions.
- sr_port/gdeerrors.msg
- Minor message change to force recreation of the associated _ctl file.
- sr_port/goframes.c
- Change some "spaces" white space to tabs.
- Change the logic where a frame expecting a return code is unwound and a value placed
such that if in a call-in and ZHALT has specified a return value, use that value
instead of the default NULL/0 value. Allows ZHALT to return its value to call-in user.
- sr_port/gtm_common_defs.h
- Remove #defines from no-longer-supported platforms.
- Nullify output of error_def() macro as it is not needed anymore as all error #defines
are always available to all routines that include mdef.h.
- sr_port/gtm_threadgbl_defs.h
- Add gtmci_retval that holds the mval address where a return value is to be set if
the frame is unwound via goframes(). Presently used to return value of ZHALT in
call-in mode.
- sr_port/mdef.h
- Add includes for ydbmerrors.h and ydberrors.h so all error codes are available to
every compilation.
- sr_port/merrors.msg
- Minor message changes to force recreation of the associated _ctl file.
- sr_port/msg.m
- Add an $ETRAP so errors aren't invisible.
- Remove creation of LITDEF entries for each error id in *_ctl.c files.
- Create ydb*.h header files that contain a #define for each error id.
- For merrors.h, also create libydberrors.h that contain a YDB_ prefixed version of the
error name (e.g. YDB_ERR_ACK) and whose value is the negated value of the actual id for
use in the libyottadb API.
- Remove VMS paths.
- sr_port/mumps.hlp
- Fix message output in a ZMESSAGE example.
- sr_port/op_halt.c
- The return code back to dm_start should be SUCCESS (1) instead of zero for a
missing return code to set correctly.
- sr_port/op_zg1.c
- Set the return code to dm_start to be SUCCESS overriding any previous value set.
- sr_port/op_zhalt.c
- Set the return code to dm_start to be SUCCESS instead of the argument value. The
argument value is saved in the gtmci_retval thread global for goframes() to find
and use.
- Return SUCCESS to dm_start().
- sr_port/ydberrors.msg
- Minor message changes to force recreation of the associated _ctl file.
- sr_unix/cmidefsp.h
- Remove #defines for CMI_REASON_{IODONE, INTMSG, CONNECT} since they duplicate definitions
in ydbcmierrors.h and have different values (don't know why).
- sr_unix/errorsp.h
- Mostly cosmetic fix but change include to use quotes instead of angle brackets on name.
- sr_unix/exttab_parse.c
- Relocate static tables from mid-code to top of routine per code standards.
- Rename all gtm_* types to ydb_* types.
- sr_unix/fgncalsp.h
- Rename all gtm_* types to ydb_* types.
- sr_unix/gtm.c
- Remove conditional code for unsupported platforms.
- Remove integer definitions for some errors as these errors are all defined by the
the include pulled in by mdef.h.
- sr_unix/gtm_env.csh
- Change gtm_ci entry point name to ydb_ci in the linker options definition.
- sr_unix/gtm_trigger.c
- Factor out mumps_status initialization this is not a valid value for mumps_status
but it serves to clear out any other (error) value that might be there as we
gear up to start the trigger.
- Added comment to mumps_status initialization to 0. This value does not last as it
is almost immediately reset to SUCCESS (1) by dm_start.
-sr_unix/gtmci.c
- Change static declaration of get_entry to a STATICFNDCL/STATICDEF definition instead.
- Rename all gtm_* types and routines to ydb_* types.
- Initialize gtmci_retval on entry to ydb_ci[p.j]().
- Verify if calling a routine with parms that it has a formal list and raise the
FMLLSTMISSING if not.
- Rearrange some of the case statemments to stack entries doing same action.
- Add call to op_exfunret() to validate/verify the return value if have one.
This raises appropriate errors if the return value is not expected or is otherwise
appropriate.
- sr_unix/gtmci.h
- Rename gtm_ci_exec() to ydb_ci_exec()
- sr_unix/gtmci_signals.c
- Code standard cosmetic changes to put opening brace on its own line.
- sr_unix/gtmexe_symbols.exp
- sr_unix/gtmshr_symbols.exp
- Change call-in related exported names from gtm_* to ydb_*
- sr_unix/gtmxc_types.h
- sr_unix/ydbxc_types.h
- Separated gtmxc_types.h into two parts:
1. ydbxc_types.h - contains the ydb_* call-in routine definitions plus the ydb_* type
definitions.
2. gtmxc_types.h - includes ydbxc_types.h and also includes both the gtm_* types
(typedef'd to ydb types) and the long deprecated xc_* types. In addition it contains
some #defines to turn gtm_* call-in routine invocations into the new ydb_* routines.
- sr_unix/op_fnfgncal.c
- Rename all gtm_* types to ydb_* types.
- sr_unix_gnp/cmierrors.msg
- Minor message changes to force recreation of the associated _ctl file.
- sr_x86_64/gdeerrors_ctl.c
- Minor message change (in .msg file) to force recreation of this _ctl file.
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_x86_64/libydberrors.h
estess
added a commit
to estess/YottaDB
that referenced
this issue
Nov 20, 2017
…commit to libyottadb.h which will hold the needed definitions for the libyottadb simple API as well as the existing call-in, call-out, plug-in etc uses of the types and routines it contains.
estess
added a commit
to estess/YottaDB
that referenced
this issue
Nov 20, 2017
…le API - part 1
This commit contains multiple (now squashed) mini-commits as described below:
Drop 11/11/2017:
- No major functionality add with this drop. The primary change was renaming
ydbxc_types.h originally created in the 10/19/2017 drop to libyottadb.h and
the minor ripples that caused.
Drop 11/08/2017:
- Change how error message ids were defined in the code base. They used to be
global integer values. I think this was to keep from having to include a
large header file with #define values for the error ids but now even ARM
processors handle this situation reasonably well.
- Also retrofitted all changes to the new assembler modules for the newly
supported ARM platform.
- Lastly, a regression or two with call-ins have been fixed and a few edge-case
day-one issues also with call-ins have been fixed:
1. HALT no longer halts but returns to the caller
2. ZHALT no longer halts but returns its argument to the caller as the
return value.
3. QUIT with no arg when a return value was requested raises an error.
4. QUIT with an arg when no return value was expected raises an error.
5. Call with args to a routine that has none defined raises an error.
6. Fixes for various issues (mostly sig-11 segmentation violations) dealing
with return values when HALT/ZHALT/ZGOTO 0 are unwinding the stack to return
either a specified return value (ZHALT) or a default NULL or zero return
value (HALT or ZGOTO 0).
Drop 10/19/2017:
- No real functionality additions. This drop is mostly "foundational" - likely
a few more of these before anything really interesting comes along but it
allows the new call-in interface to be used.
- Changes:
- All of the call-in routines have been renamed to have a ydb_ prefix
instead of a gtm_ prefix (e.g. gtm_cip() -> ydb_cip()). Note the gtm_
prefix names are still available but a macro definition in gtmxc_types.h
converts them to ydb_ calls instead.
- All of the call-in/external-call data types (e.g. gtm_int_t) have been
changed to have a ydb_ prefix. Note the gtm_ prefix is still supported as
are the native types (int, char, etc) and the long deprecated xc_ types.
Note also there is a slight performance bump for using ydb_ types (only
during initialization - no long term benefit).
- The gtmxc_types.h header file has been separated into two parts:
1. gtmxc_types.h contains the gtm_ prefixed type definitions as well as
the macros that change calls to gtm_ prefixed callin routines to ydb_
routine calls. Also in here are the deprecated xc_ type definitions.
It also includes ydbxc_types.h (note ydbxc_types.h was renamed to
libyottadb.h in the 11/11/2017 drop).
2. ydbxc_types.h contains the current set of call-in definitions with no
backlevel support.
Changes per module:
- All sr_armv7l/ assembler modules
- Had cosmetic changes like putting all assembler directives at the first tab
instead of in column zero to make things more standardized.
- All sr_armv7l/ AND sr_x86_64/ assembler modules
- The .include for debug.si is changed to a #include so the C preprocessor pulls
it in instead. This allows the ERR_GTMCHECK reference in the macro definitions
in this include to get their numeric substitutions from "ydbmerrors.h".
- sr_armv7l/ci_restart.s
- removed as no longer needed
- sr_armv7l/cmerrors_ctl.c
- sr_x86_64/cmerrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/cmierrors_ctl.c
- sr_x86_64/cmierrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/debug.si
- sr_x86_64/debug.si
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/dm_start.s
- sr_x86_64/dm_start.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/gdeerrors_ctl.c
- sr_x86_64/gdeerrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/libydberrors.h
- sr_x86_64/libydberrors.h
- New file with #defines of error ids prefixed by YDB_ and with negative values that
are used in the new simple API.
- sr_armv7l/merrors_ctl.c
- sr_x86_64/merrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/op_exfun.s
- sr_x86_64/op_exfun.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_extcall.s
- sr_x86_64/op_extcall.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_extexfun.s
- sr_x86_64/op_extexfun.s
- Change ERR_GTMCHECK & ERR_FMLLSTMISSING access from global int to #define value
- sr_armv7l/op_extjmp.s
- sr_x86_64/op_extjmp.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofexfun.s
- sr_x86_64/op_mprofexfun.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofextcall.s
- sr_x86_64/op_mprofextcall.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofextexfun.s
- sr_x86_64/op_mprofextexfun.s
- Change ERR_GTMCHECK & ERR_FMLLSTMISSING access from global int to #define value
- sr_armv7l/ydbcmerrors.h
- sr_x86_64/ydbcmerrors.h
- Remove stuff that belongs in cmerrors_ctl.c (invalid generation and botched rebase).
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydbcmierrors.h
- sr_x86_64/ydbcmierrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydberrors.h
- sr_x86_64/ydberrors.h
- Remove stuff that belongs in ydberrors_ctl.c (invalid generation and botched rebase).
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydberrors_ctl.c
- sr_x86_64/ydberrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/ydbgdeerrors.h
- sr_x86_64/ydbgdeerrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydbmerrors.h
- sr_x86_64/ydbmerrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_linux/platform.cmake
- Add specific include option for ydbmerrors.h for all assembler routines in both
armv7l and x86_64 platforms.
- Add option so all assembler routines are first processed with the C pre-processor
to pick up error id definitions and do appropriate constant substitution.
- Change the names of the entry points that we link with since all call-in entry point
names changed from gtm_ to ydb_
- sr_port/cmidef.h
- Add include to pick up new #define style cmierrors definitions.
- sr_port/cmmdef.h
- Add include to pick up new #define style cmerrors definitions.
- sr_port/goframes.c
- Change some "spaces" white space to tabs.
- Change the logic where a frame expecting a return code is unwound and a value placed
such that if in a call-in and ZHALT has specified a return value, use that value
instead of the default NULL/0 value. Allows ZHALT to return its value to call-in user.
- sr_port/gtm_common_defs.h
- Remove #defines from no-longer-supported platforms.
- Nullify output of error_def() macro as it is not needed anymore as all error #defines
are always available to all routines that include mdef.h.
- sr_port/gtm_threadgbl_defs.h
- Add gtmci_retval that holds the mval address where a return value is to be set if
the frame is unwound via goframes(). Presently used to return value of ZHALT in
call-in mode.
- sr_port/mdef.h
- Add includes for ydbmerrors.h and ydberrors.h so all error codes are available to
every compilation.
- sr_port/merrors.msg
- Minor changes in comments (GTM -> YottaDB).
- sr_port/msg.m
- Add an $ETRAP so errors aren't invisible.
- Remove creation of LITDEF entries for each error id in *_ctl.c files.
- Create ydb*.h header files that contain a #define for each error id.
- For merrors.h, also create libydberrors.h that contain a YDB_ prefixed version of the
error name (e.g. YDB_ERR_ACK) and whose value is the negated value of the actual id for
use in the libyottadb API.
- Remove VMS paths.
- sr_port/mumps.hlp
- Fix message output in a ZMESSAGE example.
- sr_port/op_halt.c
- The return code back to dm_start should be SUCCESS (1) instead of zero to prevent
ydb_ci[p]() from treating this as an error return which bypasses setting all output
values (including return values) that the call-in was expected to fill. HALT
depends on unwind processing in goframes() to supply a return value if one was
expected from this call-in.
- sr_port/op_zg1.c
- Set the return code to dm_start to be SUCCESS overriding any previous value set.
- sr_port/op_zhalt.c
- Set the return code to dm_start to be SUCCESS instead of the argument value. The
argument value is saved in the gtmci_retval thread global for goframes() to find
and use.
- Return SUCCESS to dm_start().
- sr_unix/cmidefsp.h
- Remove #defines for CMI_REASON_{IODONE, INTMSG, CONNECT} since they duplicate definitions
in ydbcmierrors.h and have different values (don't know why).
- sr_unix/errorsp.h
- Mostly cosmetic fix but change include to use quotes instead of angle brackets on name.
- sr_unix/exttab_parse.c
- Relocate static tables from mid-code to top of routine per code standards.
- Rename all gtm_* types to ydb_* types.
- sr_unix/fgncalsp.h
- Rename all gtm_* types to ydb_* types.
- sr_unix/gtm.c
- Remove conditional code for unsupported platforms.
- Remove integer definitions for some errors as these errors are all defined by the
the include pulled in by mdef.h.
- sr_unix/gtm_env.csh
- Change gtm_ci entry point name to ydb_ci in the linker options definition.
- sr_unix/gtm_trigger.c
- Factor out mumps_status initialization this is not a valid value for mumps_status
but it serves to clear out any other (error) value that might be there as we
gear up to start the trigger.
- Added comment to mumps_status initialization to 0. This value does not last as it
is almost immediately reset to SUCCESS (1) by dm_start.
-sr_unix/gtmci.c
- Change static declaration of get_entry to a STATICFNDCL/STATICDEF definition instead.
- Rename all gtm_* types and routines to ydb_* types.
- Initialize gtmci_retval on entry to ydb_ci[p.j]().
- Verify if calling a routine with parms that it has a formal list and raise the
FMLLSTMISSING if not.
- Rearrange some of the case statemments to stack entries doing same action.
- Add call to op_exfunret() to validate/verify the return value if have one.
This raises appropriate errors if the return value is not expected or is otherwise
appropriate.
- sr_unix/gtmci.h
- Rename gtm_ci_exec() to ydb_ci_exec()
- sr_unix/gtmci_signals.c
- Code standard cosmetic changes to put opening brace on its own line.
- sr_unix/gtmexe_symbols.exp
- sr_unix/gtmshr_symbols.exp
- Change call-in related exported names from gtm_* to ydb_*
- sr_unix/gtmxc_types.h
- sr_unix/ydbxc_types.h
- Separated gtmxc_types.h into two parts:
1. ydbxc_types.h - contains the ydb_* call-in routine definitions plus the ydb_* type
definitions.
2. gtmxc_types.h - includes ydbxc_types.h and also includes both the gtm_* types
(typedef'd to ydb types) and the long deprecated xc_* types. In addition it contains
some #defines to turn gtm_* call-in routine invocations into the new ydb_* routines.
- sr_unix/op_fnfgncal.c
- Rename all gtm_* types to ydb_* types.
- sr_x86_64/gdeerrors_ctl.c
- Minor message change (in .msg file) to force recreation of this _ctl file.
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_x86_64/libydberrors.h
estess
added a commit
to estess/YottaDB
that referenced
this issue
Nov 20, 2017
…laying foundation for simple API
Discussion of changes:
- No major functionality add with this drop.
- Change how error message ids were defined in the code base. They used to be
global integer values. I think this was to keep from having to include a
large header file with #define values for the error ids on (much) slower
systems but now even ARM processors handle this situation reasonably well.
- Also retrofitted all changes to the new assembler modules for the newly
supported ARM platform.
- Some changes in how call-ins work along with a few edge-case day-one issues
in a call-in environment have been fixed:
1. HALT no longer halts but returns to the caller.
2. ZHALT no longer halts but returns its argument to the call-in caller as
the return value.
3. QUIT with no parm when a return value was requested raises an error.
4. QUIT with an parm when no return value was expected raises an error.
5. Call with args to a routine that has none defined raises an error.
6. Fixes for various issues (mostly sig-11 segmentation violations) dealing
with return values when HALT/ZHALT/ZGOTO 0 are unwinding the stack to return
either a specified return value (ZHALT) or a default NULL or zero return
value (HALT or ZGOTO 0).
- All of the call-in routines have been renamed to have a ydb_ prefix
instead of a gtm_ prefix (e.g. gtm_cip() -> ydb_cip()). Note the gtm_
prefix names are still available but a macro definition in gtmxc_types.h
converts them to ydb_ calls instead.
- All of the call-in/external-call data types (e.g. gtm_int_t) have been
changed to have a ydb_ prefix. Note the gtm_ prefix is still supported as
are the native types (int, char, etc) and the long deprecated xc_ types.
Note also there is a slight performance bump for using ydb_ types (only
during initialization - no long term benefit).
- The gtmxc_types.h header file has been separated into two parts:
1. gtmxc_types.h contains the gtm_ prefixed type definitions as well as
the macros that change calls to gtm_ prefixed callin routines to ydb_
routine calls. Also in here are the deprecated xc_ type definitions.
It also includes libyottadb.h
2. libyottadb.h contains the current set of call-in definitions with no
compatibility support.
Changes per module:
- All sr_armv7l/ assembler modules
- Had cosmetic changes like putting all assembler directives at the first tab
instead of in column zero to make things more standardized.
- All sr_armv7l/ AND sr_x86_64/ assembler modules
- The .include for debug.si is changed to a #include so the C preprocessor pulls
it in instead. This allows the ERR_GTMCHECK reference in the macro definitions
in this include to get their numeric substitutions from "ydbmerrors.h".
- sr_armv7l/ci_restart.s
- removed as no longer needed
- sr_armv7l/cmerrors_ctl.c
- sr_x86_64/cmerrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/cmierrors_ctl.c
- sr_x86_64/cmierrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/debug.si
- sr_x86_64/debug.si
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/dm_start.s
- sr_x86_64/dm_start.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/gdeerrors_ctl.c
- sr_x86_64/gdeerrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/libydberrors.h
- sr_x86_64/libydberrors.h
- New file with #defines of error ids prefixed by YDB_ and with negative values that
are used in the new simple API.
- sr_armv7l/merrors_ctl.c
- sr_x86_64/merrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/op_exfun.s
- sr_x86_64/op_exfun.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_extcall.s
- sr_x86_64/op_extcall.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_extexfun.s
- sr_x86_64/op_extexfun.s
- Change ERR_GTMCHECK & ERR_FMLLSTMISSING access from global int to #define value
- sr_armv7l/op_extjmp.s
- sr_x86_64/op_extjmp.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofexfun.s
- sr_x86_64/op_mprofexfun.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofextcall.s
- sr_x86_64/op_mprofextcall.s
- Change ERR_GTMCHECK access from global int to #define value
- sr_armv7l/op_mprofextexfun.s
- sr_x86_64/op_mprofextexfun.s
- Change ERR_GTMCHECK & ERR_FMLLSTMISSING access from global int to #define value
- sr_armv7l/ydbcmerrors.h
- sr_x86_64/ydbcmerrors.h
- Remove stuff that belongs in cmerrors_ctl.c (invalid generation and botched rebase).
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydbcmierrors.h
- sr_x86_64/ydbcmierrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydberrors.h
- sr_x86_64/ydberrors.h
- Remove stuff that belongs in ydberrors_ctl.c (invalid generation and botched rebase).
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydberrors_ctl.c
- sr_x86_64/ydberrors_ctl.c
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_armv7l/ydbgdeerrors.h
- sr_x86_64/ydbgdeerrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_armv7l/ydbmerrors.h
- sr_x86_64/ydbmerrors.h
- New error definition file with #defines of error ids replacing global int vals.
- sr_linux/platform.cmake
- Add specific include option for ydbmerrors.h for all assembler routines in both
armv7l and x86_64 platforms.
- Add option so all assembler routines are first processed with the C pre-processor
to pick up error id definitions and do appropriate constant substitution.
- Change the names of the entry points that we link with since all call-in entry point
names changed from gtm_ to ydb_
- sr_port/cmidef.h
- Add include to pick up new #define style cmierrors definitions.
- sr_port/cmmdef.h
- Add include to pick up new #define style cmerrors definitions.
- sr_port/gdeerrors.msg
- Update reference to GT.M to YottaDB instead.
- sr_port/goframes.c
- Change some "spaces" white space to tabs.
- Change the logic where a frame expecting a return code is unwound and a value placed
such that if in a call-in and ZHALT has specified a return value, use that value
instead of the default NULL/0 value. Allows ZHALT to return its value to call-in user.
- sr_port/gtm_common_defs.h
- Remove #defines from no-longer-supported platforms.
- Nullify output of error_def() macro as it is not needed anymore as all error #defines
are always available to all routines that include mdef.h.
- sr_port/gtm_threadgbl_defs.h
- Add gtmci_retval that holds the mval address where a return value is to be set if
the frame is unwound via goframes(). Presently used to return value of ZHALT in
call-in mode.
- sr_port/mdef.h
- Add includes for ydbmerrors.h and ydberrors.h so all error codes are available to
every compilation.
- sr_port/merrors.msg
- Minor changes in comments (GTM -> YottaDB).
- Change NOTEXTRINSIC error message from "Quit does not return.." to QUIT/ZHALT reference
instead.
- sr_port/msg.m
- Add an $ETRAP so errors aren't invisible.
- Remove creation of LITDEF entries for each error id in *_ctl.c files.
- Create ydb*.h header files that contain a #define for each error id.
- For merrors.h, also create libydberrors.h that contain a YDB_ prefixed version of the
error name (e.g. YDB_ERR_ACK) and whose value is the negated value of the actual id for
use in the libyottadb API.
- Remove VMS paths.
- sr_port/mumps.hlp
- Fix message output in a ZMESSAGE example.
- sr_port/op_halt.c
- The return code back to dm_start should be SUCCESS (1) instead of zero to prevent
ydb_ci[p]() from treating this as an error return which bypasses setting all output
values (including return values) that the call-in was expected to fill. HALT
depends on unwind processing in goframes() to supply a return value if one was
expected from this call-in.
- sr_port/op_zg1.c
- Set the return code to dm_start to be SUCCESS overriding any previous value set.
- sr_port/op_zhalt.c
- Set the return code to dm_start to be SUCCESS instead of the argument value. The
argument value is saved in the gtmci_retval thread global for goframes() to find
and use.
- Return SUCCESS to dm_start().
- sr_unix/cmidefsp.h
- Remove #defines for CMI_REASON_{IODONE, INTMSG, CONNECT} since they duplicate definitions
in ydbcmierrors.h and have different values (don't know why).
- sr_unix/errorsp.h
- Mostly cosmetic fix but change include to use quotes instead of angle brackets on name.
- sr_unix/exttab_parse.c
- Relocate static tables from mid-code to top of routine per code standards.
- Rename all gtm_* types to ydb_* types.
- sr_unix/fgncalsp.h
- Rename all gtm_* types to ydb_* types.
- sr_unix/gtm.c
- Remove conditional code for unsupported platforms.
- Remove integer definitions for some errors as these errors are all defined by the
the include pulled in by mdef.h.
- sr_unix/gtm_env.csh
- Change gtm_ci entry point name to ydb_ci in the linker options definition.
- sr_unix/gtm_trigger.c
- Factor out mumps_status initialization this is not a valid value for mumps_status
but it serves to clear out any other (error) value that might be there as we
gear up to start the trigger.
- Added comment to mumps_status initialization to 0. This value does not last as it
is almost immediately reset to SUCCESS (1) by dm_start.
- sr_unix/gtmci.c
- Change static declaration of get_entry to a STATICFNDCL/STATICDEF definition instead.
- Rename all gtm_* types and routines to ydb_* types.
- Initialize gtmci_retval on entry to ydb_ci[p.j]().
- Verify if calling a routine with parms that it has a formal list and raise the
FMLLSTMISSING if not.
- Rearrange some of the case statemments to stack entries doing same action.
- Add call to op_exfunret() to validate/verify the return value if have one.
This raises appropriate errors if the return value is not expected or is otherwise
appropriate.
- sr_unix/gtmci.h
- Rename gtm_ci_exec() to ydb_ci_exec()
- sr_unix/gtmci_signals.c
- Code standard cosmetic changes to put opening brace on its own line.
- sr_unix/gtmexe_symbols.exp
- sr_unix/gtmshr_symbols.exp
- Change call-in related exported names from gtm_* to ydb_*
- sr_unix/gtmxc_types.h
- sr_unix/libyottadb.h
- Separated gtmxc_types.h into two parts:
1. libyottadb.h - contains the ydb_* call-in routine definitions plus the ydb_* type
definitions. Basically all the base types supported by YottaDB. It also contains the
declarations for other exported entry points callable by code outside of YottaDB.
(Note there is still re-branding work to do with some of the interface routines but
those that are primarily call-in or call-out related are taken care of by this update.
2. gtmxc_types.h - includes libyottadb.h and also includes both the gtm_* types
(typedef'd to ydb types) and the long deprecated xc_* types. In addition it contains
some #defines to turn gtm_* call-in routine invocations into the new ydb_* routines.
This is now primarily a compatibility header file in that it supports the previous
type and routine naming conventions.
- sr_unix/op_fnfgncal.c
- Rename all gtm_* types to ydb_* types.
- sr_x86_64/gdeerrors_ctl.c
- Minor message change (in .msg file) to force recreation of this _ctl file.
- Removal of the error message id LITDEFs as these are changed to #defines now.
- sr_x86_64/libydberrors.h
estess
added a commit
that referenced
this issue
Nov 20, 2017
#92) * [#59] [estess] [nars1] Changes fixing/supporting call-ins and laying foundation for simple API Discussion of changes: - No major functionality add with this drop. - Change how error message ids were defined in the code base. They used to be global integer values. I think this was to keep from having to include a large header file with #define values for the error ids on (much) slower systems but now even ARM processors handle this situation reasonably well. - Also retrofitted all changes to the new assembler modules for the newly supported ARM platform. - Some changes in how call-ins work along with a few edge-case day-one issues in a call-in environment have been fixed: 1. HALT no longer halts but returns to the caller. 2. ZHALT no longer halts but returns its argument to the call-in caller as the return value. 3. QUIT with no parm when a return value was requested raises an error. 4. QUIT with an parm when no return value was expected raises an error. 5. Call with args to a routine that has none defined raises an error. 6. Fixes for various issues (mostly sig-11 segmentation violations) dealing with return values when HALT/ZHALT/ZGOTO 0 are unwinding the stack to return either a specified return value (ZHALT) or a default NULL or zero return value (HALT or ZGOTO 0). - All of the call-in routines have been renamed to have a ydb_ prefix instead of a gtm_ prefix (e.g. gtm_cip() -> ydb_cip()). Note the gtm_ prefix names are still available but a macro definition in gtmxc_types.h converts them to ydb_ calls instead. - All of the call-in/external-call data types (e.g. gtm_int_t) have been changed to have a ydb_ prefix. Note the gtm_ prefix is still supported as are the native types (int, char, etc) and the long deprecated xc_ types. Note also there is a slight performance bump for using ydb_ types (only during initialization - no long term benefit). - The gtmxc_types.h header file has been separated into two parts: 1. gtmxc_types.h contains the gtm_ prefixed type definitions as well as the macros that change calls to gtm_ prefixed callin routines to ydb_ routine calls. Also in here are the deprecated xc_ type definitions. It also includes libyottadb.h 2. libyottadb.h contains the current set of call-in definitions with no compatibility support. Changes per module: - All sr_armv7l/ assembler modules - Had cosmetic changes like putting all assembler directives at the first tab instead of in column zero to make things more standardized. - All sr_armv7l/ AND sr_x86_64/ assembler modules - The .include for debug.si is changed to a #include so the C preprocessor pulls it in instead. This allows the ERR_GTMCHECK reference in the macro definitions in this include to get their numeric substitutions from "ydbmerrors.h". - sr_armv7l/ci_restart.s - removed as no longer needed - sr_armv7l/cmerrors_ctl.c - sr_x86_64/cmerrors_ctl.c - Removal of the error message id LITDEFs as these are changed to #defines now. - sr_armv7l/cmierrors_ctl.c - sr_x86_64/cmierrors_ctl.c - Removal of the error message id LITDEFs as these are changed to #defines now. - sr_armv7l/debug.si - sr_x86_64/debug.si - Change ERR_GTMCHECK access from global int to #define value - sr_armv7l/dm_start.s - sr_x86_64/dm_start.s - Change ERR_GTMCHECK access from global int to #define value - sr_armv7l/gdeerrors_ctl.c - sr_x86_64/gdeerrors_ctl.c - Removal of the error message id LITDEFs as these are changed to #defines now. - sr_armv7l/libydberrors.h - sr_x86_64/libydberrors.h - New file with #defines of error ids prefixed by YDB_ and with negative values that are used in the new simple API. - sr_armv7l/merrors_ctl.c - sr_x86_64/merrors_ctl.c - Removal of the error message id LITDEFs as these are changed to #defines now. - sr_armv7l/op_exfun.s - sr_x86_64/op_exfun.s - Change ERR_GTMCHECK access from global int to #define value - sr_armv7l/op_extcall.s - sr_x86_64/op_extcall.s - Change ERR_GTMCHECK access from global int to #define value - sr_armv7l/op_extexfun.s - sr_x86_64/op_extexfun.s - Change ERR_GTMCHECK & ERR_FMLLSTMISSING access from global int to #define value - sr_armv7l/op_extjmp.s - sr_x86_64/op_extjmp.s - Change ERR_GTMCHECK access from global int to #define value - sr_armv7l/op_mprofexfun.s - sr_x86_64/op_mprofexfun.s - Change ERR_GTMCHECK access from global int to #define value - sr_armv7l/op_mprofextcall.s - sr_x86_64/op_mprofextcall.s - Change ERR_GTMCHECK access from global int to #define value - sr_armv7l/op_mprofextexfun.s - sr_x86_64/op_mprofextexfun.s - Change ERR_GTMCHECK & ERR_FMLLSTMISSING access from global int to #define value - sr_armv7l/ydbcmerrors.h - sr_x86_64/ydbcmerrors.h - Remove stuff that belongs in cmerrors_ctl.c (invalid generation and botched rebase). - New error definition file with #defines of error ids replacing global int vals. - sr_armv7l/ydbcmierrors.h - sr_x86_64/ydbcmierrors.h - New error definition file with #defines of error ids replacing global int vals. - sr_armv7l/ydberrors.h - sr_x86_64/ydberrors.h - Remove stuff that belongs in ydberrors_ctl.c (invalid generation and botched rebase). - New error definition file with #defines of error ids replacing global int vals. - sr_armv7l/ydberrors_ctl.c - sr_x86_64/ydberrors_ctl.c - Removal of the error message id LITDEFs as these are changed to #defines now. - sr_armv7l/ydbgdeerrors.h - sr_x86_64/ydbgdeerrors.h - New error definition file with #defines of error ids replacing global int vals. - sr_armv7l/ydbmerrors.h - sr_x86_64/ydbmerrors.h - New error definition file with #defines of error ids replacing global int vals. - sr_linux/platform.cmake - Add specific include option for ydbmerrors.h for all assembler routines in both armv7l and x86_64 platforms. - Add option so all assembler routines are first processed with the C pre-processor to pick up error id definitions and do appropriate constant substitution. - Change the names of the entry points that we link with since all call-in entry point names changed from gtm_ to ydb_ - sr_port/cmidef.h - Add include to pick up new #define style cmierrors definitions. - sr_port/cmmdef.h - Add include to pick up new #define style cmerrors definitions. - sr_port/gdeerrors.msg - Update reference to GT.M to YottaDB instead. - sr_port/goframes.c - Change some "spaces" white space to tabs. - Change the logic where a frame expecting a return code is unwound and a value placed such that if in a call-in and ZHALT has specified a return value, use that value instead of the default NULL/0 value. Allows ZHALT to return its value to call-in user. - sr_port/gtm_common_defs.h - Remove #defines from no-longer-supported platforms. - Nullify output of error_def() macro as it is not needed anymore as all error #defines are always available to all routines that include mdef.h. - sr_port/gtm_threadgbl_defs.h - Add gtmci_retval that holds the mval address where a return value is to be set if the frame is unwound via goframes(). Presently used to return value of ZHALT in call-in mode. - sr_port/mdef.h - Add includes for ydbmerrors.h and ydberrors.h so all error codes are available to every compilation. - sr_port/merrors.msg - Minor changes in comments (GTM -> YottaDB). - Change NOTEXTRINSIC error message from "Quit does not return.." to QUIT/ZHALT reference instead. - sr_port/msg.m - Add an $ETRAP so errors aren't invisible. - Remove creation of LITDEF entries for each error id in *_ctl.c files. - Create ydb*.h header files that contain a #define for each error id. - For merrors.h, also create libydberrors.h that contain a YDB_ prefixed version of the error name (e.g. YDB_ERR_ACK) and whose value is the negated value of the actual id for use in the libyottadb API. - Remove VMS paths. - sr_port/mumps.hlp - Fix message output in a ZMESSAGE example. - sr_port/op_halt.c - The return code back to dm_start should be SUCCESS (1) instead of zero to prevent ydb_ci[p]() from treating this as an error return which bypasses setting all output values (including return values) that the call-in was expected to fill. HALT depends on unwind processing in goframes() to supply a return value if one was expected from this call-in. - sr_port/op_zg1.c - Set the return code to dm_start to be SUCCESS overriding any previous value set. - sr_port/op_zhalt.c - Set the return code to dm_start to be SUCCESS instead of the argument value. The argument value is saved in the gtmci_retval thread global for goframes() to find and use. - Return SUCCESS to dm_start(). - sr_unix/cmidefsp.h - Remove #defines for CMI_REASON_{IODONE, INTMSG, CONNECT} since they duplicate definitions in ydbcmierrors.h and have different values (don't know why). - sr_unix/errorsp.h - Mostly cosmetic fix but change include to use quotes instead of angle brackets on name. - sr_unix/exttab_parse.c - Relocate static tables from mid-code to top of routine per code standards. - Rename all gtm_* types to ydb_* types. - sr_unix/fgncalsp.h - Rename all gtm_* types to ydb_* types. - sr_unix/gtm.c - Remove conditional code for unsupported platforms. - Remove integer definitions for some errors as these errors are all defined by the the include pulled in by mdef.h. - sr_unix/gtm_env.csh - Change gtm_ci entry point name to ydb_ci in the linker options definition. - sr_unix/gtm_trigger.c - Factor out mumps_status initialization this is not a valid value for mumps_status but it serves to clear out any other (error) value that might be there as we gear up to start the trigger. - Added comment to mumps_status initialization to 0. This value does not last as it is almost immediately reset to SUCCESS (1) by dm_start. - sr_unix/gtmci.c - Change static declaration of get_entry to a STATICFNDCL/STATICDEF definition instead. - Rename all gtm_* types and routines to ydb_* types. - Initialize gtmci_retval on entry to ydb_ci[p.j](). - Verify if calling a routine with parms that it has a formal list and raise the FMLLSTMISSING if not. - Rearrange some of the case statemments to stack entries doing same action. - Add call to op_exfunret() to validate/verify the return value if have one. This raises appropriate errors if the return value is not expected or is otherwise appropriate. - sr_unix/gtmci.h - Rename gtm_ci_exec() to ydb_ci_exec() - sr_unix/gtmci_signals.c - Code standard cosmetic changes to put opening brace on its own line. - sr_unix/gtmexe_symbols.exp - sr_unix/gtmshr_symbols.exp - Change call-in related exported names from gtm_* to ydb_* - sr_unix/gtmxc_types.h - sr_unix/libyottadb.h - Separated gtmxc_types.h into two parts: 1. libyottadb.h - contains the ydb_* call-in routine definitions plus the ydb_* type definitions. Basically all the base types supported by YottaDB. It also contains the declarations for other exported entry points callable by code outside of YottaDB. (Note there is still re-branding work to do with some of the interface routines but those that are primarily call-in or call-out related are taken care of by this update. 2. gtmxc_types.h - includes libyottadb.h and also includes both the gtm_* types (typedef'd to ydb types) and the long deprecated xc_* types. In addition it contains some #defines to turn gtm_* call-in routine invocations into the new ydb_* routines. This is now primarily a compatibility header file in that it supports the previous type and routine naming conventions. - sr_unix/op_fnfgncal.c - Rename all gtm_* types to ydb_* types. - sr_x86_64/gdeerrors_ctl.c - Minor message change (in .msg file) to force recreation of this _ctl file. - Removal of the error message id LITDEFs as these are changed to #defines now. - sr_x86_64/libydberrors.h * Fixes due to review comments
nars1
added a commit
to nars1/YottaDB
that referenced
this issue
Dec 5, 2017
…by the java test Below are the issues. 1) Various modules : The prior commit changed names of functions to have ydb_ prefix instead of the gtm_ prefix. But the latest java plugin has the function names with the gtm_ prefix hardcoded and since we do not want to maintain the java plugin at this point in time, we need to provide function names with both the ydb_ and gtm_ prefix. 2) gtmci.c : The prior commit had added a call to op_exfunret() in the ydb_cij() function. But this was in the wrong place (should have been in the "if" block instead of the "else" block). This is now fixed like how it is in the ydb_ci_exec() function. 3) op_fnfgncal.c : The prior commit had changed the name "gtm_xcj" to be "ydb_xcj". But the java plugin is the one that defines this function and, again since we want to avoid having to maintain the java plugin at this point in time, the name check done here is reverted to use "gtm_xcj". The java test passes after these fixes.
nars1
added a commit
that referenced
this issue
Dec 5, 2017
Below are the issues. 1) Various modules : The prior commit changed names of functions to have ydb_ prefix instead of the gtm_ prefix. But the latest java plugin has the function names with the gtm_ prefix hardcoded and since we do not want to maintain the java plugin at this point in time, we need to provide function names with both the ydb_ and gtm_ prefix. 2) gtmci.c : The prior commit had added a call to op_exfunret() in the ydb_cij() function. But this was in the wrong place (should have been in the "if" block instead of the "else" block). This is now fixed like how it is in the ydb_ci_exec() function. 3) op_fnfgncal.c : The prior commit had changed the name "gtm_xcj" to be "ydb_xcj". But the java plugin is the one that defines this function and, again since we want to avoid having to maintain the java plugin at this point in time, the name check done here is reverted to use "gtm_xcj". The java test passes after these fixes.
nars1
pushed a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
…b_set_s() simpleAPI - Also fixes issue where the new message file ydberrors.msg was having two files created (ydberrors.h and libydberrors2.h) but these extensions of the error definitions were not chained to from the main error definition files created from merrors.msg (ydbmerrors.h and libydberrors.h respectively). These latter two files now have an include at the bottom of them that includes ydberrors.h/libydberrors2.h as appropriate.
nars1
pushed a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
nars1
added a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
…SS is done again to ensure lcl_gtm_threadgbl is updated to a non-NULL value for the caller function
nars1
pushed a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
nars1
added a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
…ydb_init() invocation This means we can invoke ydb_ci() without invoking ydb_init() like was necessary in r1.10.
nars1
pushed a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
nars1
pushed a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
…tine will be in its own module. Also separated out the error handler as that is going to get significantly larger. Added code to track the rebuffered inputs so garbage collection can get ahold of them. The lvnsetstress subtest is still intermittently failing.
nars1
added a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
nars1
added a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
…commits on top of stashed work (had to manually copy sr_unix/libyottadb_int.h and sr_unix/ydb_set_s.c)
nars1
added a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
…want this pointing to caller base C program memory which could be overwritten)
nars1
pushed a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
1. (libyottadb.h) Rename LYDB_BUFFER_LITERAL() macro to match spec's YDB_STRLIT_TO_BUFFER(). 2. (libyottadb_int.h) Modify MAX_SAPI_MSTR_GC_INDX to include slots for varname and value. 3. (libyottadb_int.h) Add separate SVNOSET error if ISV is not settable. 4. (libyottadb_int.h) Return the correct value for the ISV (not index into array but the opcode). 5. (libyottadb_int.h) Add RECORD_MSTR_FOR_GC() macro for the varname rebuffer so is tracked by GC.
nars1
added a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
… error) needs to be raised for an invalid value b) Do not return an error code from within macro after an ESTABLISH has been done in caller (causes problems with condition handler stack) but instead do rts_error just like other places do (which will automatically unwind the established condition handler)
nars1
added a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
… if they are declared as fatal type (issue was IS_GTM_ERROR macro only examined merrors_ctl and not ydberrors_ctl array)
nars1
added a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
… as first char after the ^ but not in later chars)
nars1
pushed a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
…a new error message when no buffer for value is supplied
nars1
pushed a commit
to nars1/YottaDB
that referenced
this issue
Dec 26, 2017
…dd declaration for ydb_get_s().
nars1
added a commit
that referenced
this issue
Mar 27, 2018
nars1
pushed a commit
that referenced
this issue
Mar 27, 2018
21. [DONE] Add condition handlers to utility routines.
22. [DONE] Add check for process_exiting to utility routines to prevent
calls after YottaDB closes down (ydb_exit driven).
23. [DONE] Make TPTIMEOUT work. Added outofband checking to each
simpleAPI call and blocked the other 3 (maybe 4) outofband
types if the environment is simpleAPI (current frame is simpleAPI
base frame). Also allow ^C (in outofband processing) to exit for
simpleAPI mode.
26. [DONE] Change ydb_exit() where it checks that it is not running an
external call that check should be bypassed when the process is
exiting (process_exiting is set).
28. [DONE] Change YDB_MAX_STR from (1 * 1000 * 1000) to (1 * 1024 * 1024).
nars1
pushed a commit
that referenced
this issue
Mar 27, 2018
24. [DONE] Change gtmdbglvl to ydb_dbglvl.
25. [DONE] Change gtmgbldir to ydb_gbldir.
27. [DONE] Change gtm_maxtptime to ydb_maxtptime.
29. [DONE] Do not set $ECODE (done in set_zstatus) in simpleAPI mode.
30. [DONE] Change enum delete type to int type in routine and in libyottadb.
Also change var name to deltype instead of delete_method.
31. [DONE] In ydb_delete_s() where we check the delete method, bad value
should raise UNIMPLOP instead of assertpro.
32. [DONE] Change ydb_delete_s() so if no vars are specified it is an error
instead of having it delete all vars.
33. [DONE] If no vars are specified to ydb_delete_excl_s(), then delete all
local variables.
34. [DONE] Change the parameter names to match the spec in the utility
routines.
35. [DONE] In ydb_timer_start.c need to convert the time parameter to ms
before checking it for max value and sending it along.
36. a. [DONE] Change the name of YDB_MAX_LOCKTIME to YDB_MAX_TIME to make
it more general so the max can be used elsewhere.
b. [DONE] Change the name/content of ERR_TIMEOUT2LONG to make it more
generic also (time specified exceeds maximum sort of thing).
c. [DONE] Add the maximum time checks and raising error if too big to
ydb_hiber_start(), ydb_hiber_start_wait_any() and ydb_timer_start().
37. [DONE] Create new type of function pointer "ydb_funcptr_retvoid_t" and
use it as the function pointer type for the ydb_timer_start() parm.
nars1
pushed a commit
that referenced
this issue
Mar 27, 2018
38. [DONE] In USEC_SLEEP() change * 1000 to * 1000L so result is correct
type. Also do the same in gt_timer.c line 348 or so.
39. [DONE] Change ydb_zmaxtptime to ydb_maxtptime (like TODO 27 said) and
fix test.
40. [DONE] Remove YDB_ERR_UNKNOWN error.
42. [DONE] Change LVUNDEF_OK to ERR_LVUNDEF_OK.
43. [DONE] Change YDB_MAX_NAMES to 35.
44. [DONE] Fix error with ydb_gbldir not being picked up.
45. [DONE] Add INVNAMECOUNT check and error to ydb_tp_s() and
ydb_delete_excl_s().
46. [DONE] All error checking including parameters should occur before
all locks are released in ydb_lock_s().
47. [DONE] All ydb_* functions should return either int or void - except
the one oddball ydb_malloc() which returns a void *.
48. [DONE] Rename YDB_ERROR_SEVERITY() macro to YDB_SEVERITY() and the
severity codes to YDB_SEVERITY_*.
nars1
pushed a commit
that referenced
this issue
Mar 27, 2018
…orks correctly in 32 bit architecture.
nars1
added a commit
that referenced
this issue
Mar 27, 2018
…out (if 1 second was specified, they used to wait for 1000 seconds)
nars1
added a commit
that referenced
this issue
Mar 27, 2018
…RT2 error (which is implicit TP but is different from implicit TP created by a non-TP update in M)
nars1
added a commit
that referenced
this issue
Mar 27, 2018
…subtest, tp2.c fails with cores otherwise)
nars1
added a commit
that referenced
this issue
Mar 27, 2018
nars1
pushed a commit
that referenced
this issue
Mar 27, 2018
1. [#189] Changed openlog() so it call routines get_syslog_flags which fetches the appropriate syslog flags depending on whether /dev/log exists or not 2. [#59] YDB_MAX_TIME name changed to YDB_MAX_TIME_NSEC to reflect nano-second unit of measure. 3. [#59] Fix to sapi-return_subscr_nodes.c which would not have properly reflected INSUFFSUBS when appropriate.
nars1
pushed a commit
that referenced
this issue
Mar 27, 2018
nars1
added a commit
that referenced
this issue
Mar 27, 2018
…vious commit (simpleapi/nodenext and simpleapi/nodeprev subtests pass now)
nars1
added a commit
that referenced
this issue
Mar 27, 2018
…timer_start/ydb_hiber_start/ydb_hiber_start_wait_any to return int (instead of void) as they can return YDB_ERR_TIME2LONG
nars1
added a commit
that referenced
this issue
Mar 27, 2018
nars1
added a commit
that referenced
this issue
Mar 27, 2018
…NTTOOHI, nix MISSINGVARNAMES, MISSINGVARNAME) 1) MISSINGVARNAMES was an unused error message so removed it. 2) MISSINGVARNAME was being issued in ydb_delete_s() if input varname is NULL. But not sure why that should be treated any different than a NULL varname passed to ydb_set_s(), ydb_get_s() etc. Therefore removed that error message. 3) Renamed TOOMANYVARNAMES error name to NAMECOUNTTOOHI to be more in line with the INVNAMECOUNT error. As part of that, enhanced NAMECOUNTTOOHI to indicate the ydb_*_s() call which got that error (just like the INVNAMECOUNT error). 4) Added the NAMECOUNTTOOHI error in ydb_lock_s() (missed out in prior commit).
nars1
added a commit
that referenced
this issue
Mar 27, 2018
nars1
added a commit
that referenced
this issue
Mar 27, 2018
nars1
pushed a commit
that referenced
this issue
Mar 27, 2018
…ROR2) 1. First, the FATALERRORx messages were switched where they were being generated. So you'll see the messages reversed now. Previously FATALERROR2 (no core) was coming out when we WERE generating a core. When we generate a core, we are now putting out FATALERROR1. 2. I wasn't aware where we were generating the core but it turns out it was in ch_cond_core() which is called by the DRIVECH() macro before it drives the top condition handler. I added a !IS_SIMPLEAPI_MODE in there so we decide whether to core or not in our condition handler. Core generation for ERR_MEMORY and ERR_STACKOFLOW will probably be removed from ch_cond_core() altogether when the FATALERRORx scheme migrates to the rest of YottaDB. 3. The entire section that stops the process in ydb_simpleapi_ch() was being bypassed by the inclusion of !SUPPRESS_DUMP - which because ch_cond_core() created a core meant the handler returned to ydb_set_s() and loops like crazy. 4. Had a '!' when I should not have on the debug flag test which caused the wrong message (and lack of core once I stopped ch_cond_core() from generating one). 5. Also changed the GTM_FATAL_ERROR*.txt file to a YDB_FATAL_ERROR*.txt file.
nars1
added a commit
that referenced
this issue
Mar 27, 2018
nars1
added a commit
that referenced
this issue
Mar 27, 2018
nars1
added a commit
that referenced
this issue
Mar 27, 2018
nars1
pushed a commit
that referenced
this issue
Mar 27, 2018
1. Fix ydb_exit() to properly allow a call with process_exiting set so long as gtm_startup_active is also set. 2. Change all SKIP_BASE_FRAME invocations to SKIP_BASE_FRAMES and make the new macro be able to unwind past adjacent base frames of same or dissimilar types. 3. Visit all places where SFT_CI is used to make sure they are able to jump both TRIGGER and call-in base frames if that is appropriate (additional uses of SKIP_BASE_FRAMES was used in this effort). 4. Reverted gtm_filename_to_id() to what it was before (logic wise) undoing the attempt to return an error code. 5. Modified ydb_file_name_to_id() to no longer be a wrapper for gtm_filename_to_id() but now has simliar logic in that DOES return an error code. 6. The -1 return codes from ydb_thread_is_main() and ydb_file_is_identical() are changed to YDB_NOTOK which is an error code added to libyottadb.h (not a real error - just a value like YDB_LOCK_TIMEOUT). 7. Added invocation to create the YDB_FATAL_ERROR file when a fatal error is encountered in simpleAPI.
chathaway-codes
pushed a commit
that referenced
this issue
May 8, 2019
…_incr_s.c introduced in [#59] There was an assert in ydb_incr_s.c which expected that the result of an op_add() will never have the MV_STR bit set. But it is possible if the input node is an lvn and is not defined and the increment is a string literal mval (MV_STR bit is already set), the result of the ydb_incr_s() would be a copy of the string literal mval with the MV_STR bit set. That would fail the assert (the recently enhanced go/random_walk subtest failed this assert). Note that SET_YDB_BUFF_T_FROM_MVAL invoked at the end of ydb_incr_s() does a MV_FORCE_STR of the result if it does not already have the MV_STR bit set so the assert served no purpose and is now removed..
chathaway-codes
pushed a commit
that referenced
this issue
May 13, 2019
The r126/random_walk subtest failed with a STACKCRIT error. Analysis of this led to an M-stack leak from $QUERY(lvn,-1). Commit (SHA 10d8235) as part of #59 added a push_v1 variable but forgot to initialize it to TRUE in the case where v1 and v2 were being pushed onto the M-stack. Similar code exists in op_fnquery.c but push_v1 is correctly initialized there. The same fix is now done in op_fnreversequery.c.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Final Release Note
The YottaDB engine is directly accessible using a C-callable API. See https://docs.yottadb.com/MultiLangProgGuide/index.html for user documentation.
Description
To access data in YottaDB without this enhancement requires either writing a small shim in M or using the GNP protocol over TCP (which does not support access to local variables or transaction processing). YottaDB needs direct, in-process calls from C (and since C is the lingua Franca of computer languages, from any language that has a C interface).
Draft Release Note
The YottaDB engine is directly accessible from C. See https://github.com/ksbhaskar/YottaDBdoc/blob/master/MultiLangProgGuide.rst for the [draft] user documentation.
The text was updated successfully, but these errors were encountered: