patch: sanitizer summaries are redundant, don't emit them
richard at metafoo.co.uk
Wed Oct 23 11:22:10 PDT 2013
On Wed, Oct 23, 2013 at 12:51 AM, Kostya Serebryany <kcc at google.com> wrote:
> Okay, that works. But there's one other difference Richard brought up:
>> ReportErrorSummary expects complete error text, while printing is assumed
>> to be continually appending. UBSan manages to print out each piece of the
>> error individually and never allocates any buffers.
>> I could change UBSan's Printf calls to instead keep appending to an
>> internal buffer and then feed the buffer to
>> __sanitizer_report_error_summary when we're done. I just wanted to point
>> out the difference in these APIs, in case you can think of a way we could
>> do this without require a buffer somewhere.
> asan/tsan/msan also prints most of it's stuff directly (although there is
> in fact a buffer hidden inside Printf). It only uses an explicit buffer to
> print the summary.
> On a side note: why ubsan doesn't print the stack trace?
See the other thread: sanitizer_common's backtrace functionality currently
requires symbol interposition, which ubsan deliberately avoids.
> I can imagine cases where this will help diagnose a bug.
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cfe-commits