894: Fix error-reporting crash in non-blocking connection. #895
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
Fixes: #894 If construction of a non-blocking `connection` failed for a reason _other_ than an out-of-memory error, then reporting of the error would fail. This happened because the error-reporting code tried to read the error message _after it had already cleared the connection pointer._ This means reading deallocated memory.
In clang 19, all of a sudden the compile-time feature check for `[[assume()]]` support started issuing a warning: it says that attribute is a "C++23 extension." And the maintainer-mode build treats warnings as wrrors, so... Kaboom. The problem is that in this mode, clang 19 does define the feature test macro for this attribute even in C++ 17, but when you actually _use_ the attribute, clang warns (at least with strict checking options) that it is not part of C++17 (or C++20, as the case may be). To get around that, I wrote a more substantial feature check and ran it at _configure_ time. This is where my work on systematising the feature check mechanism really pays off — a few years ago this would have been a fairly substantial change, with lots of opportunities to do it wrong.
For some reason the line in this unit test where we create a `std::vector<pqxx::binarystring>` triggers a deprecation warning, even though the line is bracketed in an "ignore deprecations" block. Since the whole class is deprecated, it wasn't really worth fixing properly. I'll just stop building this test on clang, until the time comes to drop the whole class and its tests anyway. Which I hope is not too far off.
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Outdated suggestions cannot be applied.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.
Fixes: #894
If construction of a non-blocking
connectionfailed for a reason other than an out-of-memory error, then reporting of the error would fail. This happened because the error-reporting code tried to read the error message after it had already cleared the connection pointer. This means reading deallocated memory.There was an unrelated complication. In maintainer mode (extra-strict compiler options), the
binarystringunit test elicits a warning from clang:binarystringis deprecated. Yeah duh. There shouldn't be a warning because that piece of test code (like many others) is bracketed in "ignore deprecation" directives. In just this one case, for just this one compiler, that doesn't seem to do the trick. In the end I just disabled that one test for that one compiler. Hate to do it, but this is deprecated code so it's going away anyway.Oh, and the build also broke because clang 19 in C++17 and C++20 enables the feature test macro for the
assumeattribute... but then issues a warning in maintainer mode when you actually use the attribute. Which again in maintainer mode means that the build fails.