Linux Port WIP #295
Draft
Linux Port WIP #295
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…the linux layer [metagen_w32] Fixed: Don't modify file creation time This isn't really a big deal but I feel like ist not the intention of os_file_set_times and its also just bad form in general to modify access times beacuse it obliterates the very limited context the OS will save about our files. Feel free to revert. [metagen_linux] Added: Auto allocation of 1GB huge_pages on os_reserve_large I don't know if we actually want this or not but I figured I would at least throw it out there because its not unusual to run up multi-gigabyte debug data on very complex projects so its absolutely a thing that would be worth looking into for performance / OS-level fragmentation and also crash resistance [metagen_strings] Added: Function to find first-occurance of a substring 'str8_match_substr' This returns a range of where the match was found. [metagen_linux] Added: Dependencies 'libpthread libuuid' [metagen_linux] Fixed: Leftover bool usage was changed to B32 and int usage NOTE: Nothing has been tested at this point, too annoying, too many missing functions, hard to test when it doesn't build
…the linux layer Documentation: Changed some links to be on their own comment line. This makes it easier to copy and discern there's a usable link present Added: Implementation for os_id_from_file from TODO TODO(nick): querry struct stat with fstat(2) and use st_dev and st_ino as ids Added: Implementation for os_full_path_from_path from TODO TODO: realpath can be used to resolve full path Fixed: Some erroneous dot (.) and arrow (->) notation Added: Some helper typedefs for long and hard to remember Linux-specific types
Added: Linking for pthread, dl, and rt Tweak: Changed RAD_BUILD_ALL to just build_all. Should be more consistent with other variables now and is documented Fixed: Random pushd and pwd commands echoing unwanted directories Documentation: Added some basic comments / thoughts about the build system
Focused mainly around memory allocations this time [Metagen Linux] Fixed: Wrong lnx function name call [Build System] Fixed: Wrong syntax for calling metagen.exe in current directory
[Metagen Linux] Fixed: os_commit should now be backed by physical memory like in Windows
Removed: Hugepage support temporarily, it can't function without additional work Fixed: Potential edge case where munmap doesn't sync to disk Tweak: Try to populate all mmap calls on either commit or view open for performance Maintenance: Renamed lnx_fstat and lnx_date to LNX_fstat to be closer in line with other types Fixed: Potential fix for setting up condition variable in wrong order [Build Batch] Added: Debug environmental variable for skipping running metagen NOTE: This is mostly because I didn't want to deal with erroring and crashing programs but wanted to still get compiler output for metagen.
…mpiles and runs I haven't verified how well much of the code works but the one thing metagen uses, the OS filter iterator, seems happy and I checked a good few of the values it spat out. The only thing left is the condition variable broadcast and a few long winded things I don't know how to compelte the implementation on. I don't really know what a condition variable is or what a broadcast for it is so moving to testing other things first. Also the OS_SystemPath_ModuleLoad is not done. Maintenance: Sorted include lines- because nice. [Build Batch] Fixed: Exiting with fail status (1) on successful compile instead of success
[Build Batch] Fixed: All targets specified should build like it's supposed to in the windows version Fixed: metagen section exiting script early [Third Party] [rad_lzb_simple] Fixed: RR_ASSERT did not support clang traps which was preventing compiles Fixed: Using standard breaking label (done:) syntax without a statement (done: statement;) RANT: So aparently this entire time MSVC and gcc were using incompatible standard syntax for the label syntax, notice how I said syntax not statement. Because goto is not a staement- it's a "labelled-staement" whatever that means, which is just ever so slightly different to a regular statement, which is ended by a semicolon (;), gcc and clang (as far I know) just auto-completes the statement. And this is your daily reminder that the C/C++ Standards Commitee are stupid and failed to create a stable and portable language, and you should never take anything for granted unless its explicitly stated in the standard. Have. Fun. Programming. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf [Metagen Linux] Removed: 'pthread.so' load, it was a mistake the compiler sets it up automatically Fixed: Maybe fix for 'lnx_initial_path' not being setup properly or having integer under/overflow [Raddbg Linux] Removed: 'pthread.so' load, it was a mistake the compiler sets it up automatically Fixed: Maybe fix for 'lnx_initial_path' not being setup properly or having integer under/overflow Removed: Depreceated os_get_command_line_arguments it seems to be unused and the API changed. [Base Strings] Added: str8_substring_match from metagen
…usly Fixed: 'OS_AccessFlag_Shared' not being valid in this context Added: Minor assert for 'lnx_entity_from_handle' [Metagen Linux Linux] Fixed: 'tctx_release()' being removed possibly erroneously [Build Batch] Fixed: Missing quoting for git hash
Fixed: '__VA_ARGS' on various macros erroring with zero arguments. So use '##__VA_ARGS' extension instead. [DF GFX] Fixed: Superflous OS-specific 'Sleep()' function [Build Batch] Added: Erorr-by-default for invalid memory ordering
[Core Linux] Fixed: Slightly different os_set_large_pages semantics to win32 API Fixed: 'actual_size_out' not being written to with ring buffers Fixed: 'os_machine_name' being completely broken NOTE: The logic looks sort of correct but a return code was interpreted wrong, and the rest was completely irrelevant, the hostname spec caps to 253 ASCII chars precisely I believe. So I just replaced nearly the whole function with a minimal and working version Fixed: 'lnx_environment' not being set properly because of Linux's unspecified virtual filesize Fixed: 'os_file_open' Typo Fixed: 'os_file_open' file descriptor mode being in decimal '660' when it should be octal '0600' Added: Some "Zero-Is-Initialization" style error handling for some things Maintenance: Tagged some functions as out of date/stale/needing review for validity Fixed: Several asorted unspecified fixes for Linux platform layer [Misc] Added: Some missing from metagen base_string
[OS Linux] Added: Rudimentary thread name setting Documentation: Added 'Here Be Dragons' warning Fixed: 'os_launch_process' Fixed: Some assorted fixes and documentation/note tweaks
…es to get linux build going [OS Linux] Added: Missing implementation for 'os_properties_from_file_path' Fixed: Accidentily leftover debugging code Added: os_condition_variable_broadcast_ Removed: Some leftover debug stuff Added: Misisng main function [Font Freetype] Added: Stubbed out files [Demon Linux] Added: Stubbed out files Added: Implimented some functionality [Build Script] Added: freetype2 directory to include path Added: freetype2 dynamic link [Documentation] Documentation: Wrote some notes with compiler flags [Random] Fixed: Clang complainting about unrelated B32 and U32 atomic types
Added: stb_image (could be temporary, not important right now) [OS Core] Fixed: Missing include guards [Build Script] Fixed: Erroring on a logging command Added: Some logging to the build setup
Fixed: Added a zero-is-valid case for 'os_file_write'
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Hiya, good day,
I wanted to make a small Linux debugger and saw you open sourced a half decent debugger and thought I'd see how far I could get porting it as a hobby instead of torturing myself. I've done a big chunk now and saw you're doing little proddings on the Linux version so I thought I'd share what I have so far.
It's in an extremely unstable and tempremental state right now but a lot of it shouldn't be too stupid and I tried to comment as much findings as possible. So you can look over it for insight if you wish and whatever but **please be warned ** I have not had a chance to test hardly anything yet properly because I wanted some kind of minimum-working-set to play with.
I tried to look over some of the other pulls/issues for other people that worked on it and it seemed there was not much progress there so that's a shame... One had mostly just basic stub outs and bugfixes and the other had a minimal OpenGL API although I'm not super sure is the target you want long term because of how janky OpenGL device selection is.