This is the mail archive of the
newlib@sourceware.org
mailing list for the newlib project.
Re: [PATCH v2] Improve performance of strstr
On 10/3/18 11:06 AM, Wilco Dijkstra wrote:
Eric Blake wrote:
Passing the glibc testsuite isn't all that reassuring, as they just
barely had to squash a strstr bug this year that was not caught by their
testsuite at the time:
It's better than nothing given newlib doesn't have any testsuite at all...
Do you know of a proper test for strstr by any chance?
Nothing that was written from the grounds of code coverage, but between
the glibc test, the gnulib test [1] (which was also recently enhanced to
catch the glibc failure), and your testing (if you want to make that
public), we're probably doing fairly well. A testsuite written for 100%
coverage as determined by gcov, or performed by a fuzzer that aims to
get to the same results, might be even more reassuring; but I'm not
asking you to tackle that. I'm just pointing out that a passing test
does not always imply a passing implementation, if the test was not
written with exact knowledge of every possible branch in the code.
[1] https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org