| Summary: | super-optimized strcasecmp() false positive | ||
|---|---|---|---|
| Product: | [Developer tools] valgrind | Reporter: | Dmitry Djachenko <dimhen> |
| Component: | memcheck | Assignee: | Julian Seward <jseward> |
| Status: | RESOLVED FIXED | ||
| Severity: | minor | CC: | dimhen, jakub, pavelo, tom |
| Priority: | NOR | ||
| Version: | 3.7 SVN | ||
| Target Milestone: | --- | ||
| Platform: | Unlisted Binaries | ||
| OS: | Linux | ||
| Latest Commit: | Version Fixed In: | ||
| Attachments: |
add support for strcasecmp/strncasecmp
valgrind-3.6.0-strcasecmp.patch |
||
Yeah, this is quite severe problem in valgrind against recentish glibcs. +/* ??? STRCASECMP(VG_Z_LD_LINUX_X86_64_SO_2, strcasecmp) */ +/* ??? STRCASECMP(VG_Z_LD64_SO_1, strcasecmp) */ is unnecessary, neither strcasecmp nor strncasecmp is ever present in ld.so. That said, the patch should also handle strcasecmp_l and strncasecmp_l, which will need some configure magic to detect tolower_l/strcasecmp_l. This could be also solvable using suppressions, but then we wouldn't diagnose invalid uses of strcasecmp by the apps. What do the _l variants do? They don't seem to have any manual page on my F14 systems... Created attachment 53355 [details]
valgrind-3.6.0-strcasecmp.patch
Updated patch to also handle strcasecmp_l and strncasecmp_l.
Looks like we crossed - I just committed an extended version of the patch as r11478. *** Bug 282979 has been marked as a duplicate of this bug. *** |
Created attachment 53319 [details] add support for strcasecmp/strncasecmp Version: 3.7 SVN OS: Linux Reproducible: Didn't try Actual Results: ==23300== Use of uninitialised value of size 8 ==23300== at 0x33D63277E4: __strncasecmp_l_sse42 (strcmp.S:402) Expected Results: no diagnostique Fedora14/x64 gcc-4.5.1