- User Since
- Aug 21 2015, 4:29 PM (230 w, 14 h)
Fri, Jan 10
@yln The description of this change needs improving. The title probably ought to be "[TSan] The "thread_terminate" event is always delivered on the terminating thread" and then that should be followed with a brief justification along with a radar number (if there is one).
Fri, Dec 20
Dec 16 2019
LGTM except the possible improvement to the error message.
Dec 13 2019
Dec 12 2019
Dec 3 2019
Nov 29 2019
The only thing I'm not sure about is whether to create a new subsection for it, or just leave it as inline text and link to the paragraph directly (which is what I did).
Nov 22 2019
@yln It's still useful to know which tests are marked as UNSUPPORTED without actually running the tests. Does this change break that workflow?
Nov 19 2019
Nov 18 2019
@yln I've done another pass over this. Hopefully there are no remaining issues.
- Update some test comments.
- Use more sophisticated regex so we can drop separate 0x0 bad address check.
Nov 15 2019
@arphaman Nice catch!
Nov 14 2019
- Drop some comments. There are comments on internal_snprintf that explain the same thing.
- Use internal_snprintf in FormatString. It handles the nullptr case already.
Nov 13 2019
- Fix line length violation.
- Fix broken test.
@yln I've re-worked this patch addressing everything I could.
- Rework test to use a single binary
- Rework logic to only set function_offset when a useful value was found.
- Clean up test case to not use compiled in default options.
- Adopt a few testing tweaks suggested by @yln offline.
- Refactor string printing code into helper method to reduce code duplication.
Nov 12 2019
Nov 11 2019
- Fix MSan compilation error.
Nov 7 2019
Put in a proper defensive check rather than "defensive programming" so that we crash if someone screws up.
- Go back to a small buffer for printing values. Mmap is probably overkill here.
- Remove defensive check.
- Add trunation test.
- Style clean up in FlagHandler<const char *>::Format(char *buffer, uptr size).
- Style clean up in FlagHandlerKeepGoing.
Nov 6 2019
Nov 5 2019
- Change path name in test.
- Rename CurrentValueAsString to Format.
- Implement support for FlagHandlerKeepGoing.
- Use Mmap to allocate memory for string representation of option value.
- Switch to using internal_strlcpy instead of internal_strncpy.
- Remove separate test for include_if_exists
Nov 2 2019
Nov 1 2019
- CurrentValueAsString return code now indicates truncation
- Print "Truncated" if printed current value was truncated.
- Clean up CurrentValueAsString implementations.
- Add Support in FlagHandlerInclude for showing current value (not quite right but good enough).
@kongyi Hmm we have both x86_common/floatdixf.c and x86_64/floatdixf.c. I think both x86_64h_SOURCES and x86_64_SOURCES will end up with both x86_common/floatdixf.c and x86_64/floatdixf.c due to...
Oct 31 2019
LGTM. Thanks for the patch.
Oct 30 2019
Thanks for the review committed as 8ea148dc0cbff33ac3c80cf4273991465479a01e
Oct 29 2019
Note there is a way to modify this patch so that it does actually show the defaults in the .inc files. This would require adding storage for the default values and assigning to it in CommonFlags::SetDefaults() (and all other similar sanitizer flag parsers). Then the CurrentValueAsString() functions could be renamed to DefaultValueAsString() and could be modified to read the default value rather than the current value.
Oct 28 2019
Oct 25 2019
Thanks for the patch. I didn't know that this information was missing from our docs!
Oct 23 2019
Oct 18 2019
Fix for powerpc issue is now up for review: https://reviews.llvm.org/D69189
Okay I figured out the issue. The problem is with the filtering code. It assumes that architecture specific specializations have a filepath like <arch>/filename. The problem here is that for powerpc64 the <arch> is powerpc64 but the directory name used is actually ppc. I've got a fix for this which should prevent people from making the same mistake in the future.
Oct 17 2019
Hmm this code looks a little dubious. It's modifying a list while iterating over it and I'm not sure that regex is right.
and then later
Landed workaround to downgrade fatal error to warning in r375162.
Damn looks like something is broken for linux powerpc