glibc 2.25 seems to have broken AddressSaniitzer Inbox x glibc x sanitizer x

[Joseph Myers]

On Wed, 9 May 2018, Zack Weinberg wrote:

> I wonder how far we could get with black-box testing.  If the main
> glibc testsuite compiled and ran some programs with ASan enabled,

You'd need to have a configure test for whether the compiler includes 
libsanitizer support.  The tests would also need somehow to be structured 
so that if an address space limit is set (on the system running the tested 
glibc, not necessarily the same as the system running the compilers) they 
end up as UNRESOLVED rather than FAIL (ASan requires a very large amount 
of virtual address space).

> discussion?  It might be redundant to what a buildbot that ran the
> entire testsuite under ASan instrumentation would accomplish, but it

Running the whole glibc testsuite under ASan might be hard anyway; it's 
ASan's own tests I expect would be useful for more thorough testing with 
new glibc.

A significant part of testing, which does however require running a bot, 
would be that GCC keeps building, on architectures that support 
libsanitizer, with the new glibc headers, since changes to headers are a 
common cause of libsanitizer build breakage (as it depends on 
implementation details of headers that are not public interfaces).

-- 
Joseph S. Myers
joseph@codesourcery.com