Add support for driving BufferedProtocol instances using sock_recv_into
#60
Conversation
ntninja
force-pushed
the
feat-buffered-protocols
branch
from
2e48902
to
5aee5a5
Compare
ntninja
force-pushed
the
feat-buffered-protocols
branch
from
1ffbc38
to
c901d10
Compare
ntninja
force-pushed
the
feat-buffered-protocols
branch
from
c901d10
to
87a14a5
Compare
…_into`
Support for this feature is emulated for the `Pipe{Read,Write}Transport`
however, since `glib.IOChannel` does not appear to expose anything similar
to the `recvinto` syscall in its APIs. The emulation at least makes it possible
to use `BufferedProtocol` instances without causing crashes due to API
differences versus the classic `Protocol` interface at least.
ntninja
force-pushed
the
feat-buffered-protocols
branch
from
87a14a5
to
74a3ec4
Compare
src/gbulb/transports.py
Outdated
| assert isinstance(data, bytes) | ||
| if data != b"" and data != 0: | ||
| if self._alloc_read_buffers: | ||
| #FIXME: GLib does not actually expose the equivalent to |
There was a problem hiding this comment.
Note: This line caused flake8 to cry with error E265, because there is no space between the hash sign and the word FIXME…
|
(CI passes now, but I had to disable the broken towncrier check.) |
There was a problem hiding this comment.
Thanks for the PR (and apologies for the delay in reviewing; I apparently didn't have this repo on my Github Watch list).
I've corrected the problem with the Towncrier configuration; and this PR looks reasonable from a first-pass review. However, in order to be merged, it requires tests for the new functionality.
There was a problem hiding this comment.
I'd feel a lot better if this had tests; but empirical testing shows that it resolves #116, and the perfect is the enemy of the good, so I'll merge it. Thanks for the PR; apologies for taking so long to get to merging it.
Support for this feature is emulated for the
Pipe{Read,Write}Transporthowever, sinceglib.IOChanneldoes not appear to expose anything similar to therecvintosyscall in its APIs. The emulation at least makes it possible to useBufferedProtocolinstances without causing crashes due to API differences versus the classicProtocolinterface at least.Refs #58
Fixes #116
PR Checklist: