bpo-37705: Improve the implementation of winerror_to_errno() - #15623
zooba merged 1 commit into
Conversation
Sorry, something went wrong.
winerror_to_errno() is no longer automatically generated. Do not rely on the old _dosmapperr() function. Add ERROR_NO_UNICODE_TRANSLATION (1113) -> EILSEQ.
Sorry, something went wrong.
There was a problem hiding this comment.
Thanks for the PR @ZackerySpytz.
I am very much in favor of using constants to give the winerrors more semantic meaning and grouping the returns. However, there's several of the constant names which don't describe their purpose in a meaningful way. I'm not familiar with our stance on linking to external documentation within code comments, but it seems like it might be helpful to provide a link to https://docs.microsoft.com/en-us/windows/win32/winsock/windows-sockets-error-codes-2.
Sorry, something went wrong.
|
Stever Downer (@zooba) mentioned adding a dict in Modules/errnomodule.c that exposes this mapping to Python code. I guess it could be called winerror2errno. Maybe we could also add a winerrorcode dict with the symbolic names of mapped errors, and errors of interest. For the latter, I think it would be better if winerror_to_errno were based on a counted or null-terminated table instead of a switch. Each entry would have the symbolic name as a string (e.g. "ERROR_FILE_NOT_FOUND"), the error code, and the mapped errno value. This avoids needing to maintain supported error codes in two places. Otherwise if we map a new error in PC/errmap.h, we'll have to separately add the symbolic name in Modules/errnomodule.c. For Winsock errors, I think the winerror2errno and winerrorcode dicts would only need the six that are specially mapped to CRT errno values, since all others are already available as errno values -- both as module constants and in the errorcode dict. Given that, should these six Windows errors remain defined as errno values? For example, now that WSAEINTR maps to EINTR, does it make sense to have errno.WSAEINTR == 10004 and errorcode[10004] == 'WSAEINTR'? In any case, I don't think it's a problem to leave them there. |
Sorry, something went wrong.
|
I'm merging and backporting this, as the additional error codes are general goodness that will improve all active versions. Adding a new public API should be only for master. |
Sorry, something went wrong.
|
Thanks @ZackerySpytz for the PR, and @zooba for merging it 🌮🎉.. I'm working now to backport this PR to: 3.7, 3.8. |
Sorry, something went wrong.
|
GH-15749 is a backport of this pull request to the 3.8 branch. |
Sorry, something went wrong.
|
GH-15750 is a backport of this pull request to the 3.7 branch. |
Sorry, something went wrong.
winerror_to_errno() is no longer automatically generated. Do not rely on the old _dosmapperr() function. Add ERROR_NO_UNICODE_TRANSLATION (1113) -> EILSEQ. (cherry picked from commit 19052a1) Co-authored-by: Zackery Spytz <zspytz@gmail.com>
winerror_to_errno() is no longer automatically generated. Do not rely on the old _dosmapperr() function. Add ERROR_NO_UNICODE_TRANSLATION (1113) -> EILSEQ. (cherry picked from commit 19052a1) Co-authored-by: Zackery Spytz <zspytz@gmail.com>
…H-15623) winerror_to_errno() is no longer automatically generated. Do not rely on the old _dosmapperr() function. Add ERROR_NO_UNICODE_TRANSLATION (1113) -> EILSEQ.
…H-15623) winerror_to_errno() is no longer automatically generated. Do not rely on the old _dosmapperr() function. Add ERROR_NO_UNICODE_TRANSLATION (1113) -> EILSEQ.
…H-15623) winerror_to_errno() is no longer automatically generated. Do not rely on the old _dosmapperr() function. Add ERROR_NO_UNICODE_TRANSLATION (1113) -> EILSEQ.
After pythonGH-15623 deleted `generrmap.c`, a related mak-file stopped working. The mak contains generrmap-related rules only so it should be removed altogether. Further search for `errmap\.mak|generrmap` regex through content of CPython files shows no dangling reference left.
After GH-15623 deleted `generrmap.c`, a related mak-file stopped working. The mak contains generrmap-related rules only so it should be removed altogether. Further search for `errmap\.mak|generrmap` regex through content of CPython files shows no dangling reference left. Since generrmap is already effectively removed, this pull request contains no blurp.
winerror_to_errno() is no longer automatically generated.
Do not rely on the old _dosmapperr() function.
Add ERROR_NO_UNICODE_TRANSLATION (1113) -> EILSEQ.
Many thanks go to Eryk Sun for their posts on the BPO issue.
https://bugs.python.org/issue37705