This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: libsanitizer merge from upstream r196090


On Tue, Dec 3, 2013 at 6:53 AM, Konstantin Serebryany
<konstantin.s.serebryany@gmail.com> wrote:

>>> >> with #if LINUX_VERSION_CODE >= 132640
>>> Good idea, let me try that.
>>
>> Had a quick look at this on RHEL 5.
>> Following patch let me compile at least the first source file, but then
>> I run into tons of issues in sanitizer_platform_limits_posix.cc.
>
> That's what I am afraid of. Even if we manage to compile everything,
> there is no guarantee that the code will work.
> I suggest to simply disable libsanitizer build on the older systems
> which is what happens de facto now.
> If there is significant interest in maintaining asan&co on older
> systems (which I have not seen so far),
> then those interested will need to help us in upstream repository (llvm) by
> a) sending us patches using http://llvm.org/docs/Phabricator.html and
> b) setting up a public buildbot (attached to the LLVM master bot) with
> the system they care about.
> If there is no one interested enough to do a) and b) I say we should
> not spend time on this.

IMO, it is also OK for the configure to check for needed features and
disable libsanitizer (perhaps with some informative message) if
minimum requirements are not met. If someone adds workarounds for
those missing features to support older systems, then chese checks can
easily be adapted. The problem ATM is, that gcc won't build
out-of-the-box on older distributions, although adding
--disable-libsanitizer manually works OK.

> And this discussion does not affect the merge since nothing that works
> today will get broken, right?

Yes, that's right.

Uros.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]