- User Since
- Apr 18 2018, 2:22 AM (91 w, 3 d)
Thu, Jan 16
Wed, Jan 15
This was meant as a post-commit review
Tue, Jan 14
Fri, Jan 10
Thu, Jan 9
(and a delete)
Oh, I will just add some ASSERT to the test and a comment on what it actually tests.
Ok, seems such test fails reliably (on macOS).
Wed, Jan 8
calculate the size of the buffer upfront
I tried but couldn't reproduce a segfault. Do you have any suggestion on how to reasonably reliably (TM) reproduce it?
Dec 12 2019
Dec 4 2019
The use of xcrun looks sound to me.
Dec 2 2019
Hi @arthurp, I can review the libclang part of the patch.
Nov 21 2019
Nov 19 2019
Ahh, right. I originally wanted to support -ivfsoverlay in Driver as it seemed reasonable to check the existence of blacklist in Driver and open it in cc1 using the same fs. I got talked out of it and I didn't touch the Driver but seems like I should've replaced the calls to native fs with VFS to keep things consistent. Thanks for picking this up!
Nov 8 2019
Nov 7 2019
@pcc Could you please take a look?
I am in no particular rush and happy to help you as much as I can but at some point not too far in the future I just want this to land and not get reverted.
Nov 6 2019
The test seems to be fine (fails on master, passes with the patch). I most probably just forgot to run it.
Actually use the VFS that got passed.
Oct 31 2019
I see what you mean - I didn't realize I could use the real fs as the default arg.
I still feel that handling files is out of scope for SpecialCaseList but I don't think those couple lines are a big deal either way - updating the patch now.
Added a cc1 command line option -fsanitize-system-blacklist so driver can tell cc1 which blacklists are user-specified and which are system ones.
@eugenis Does this make your workaround obsolete?
Oct 30 2019
Oct 29 2019
Oct 21 2019
I am trying to solve two different things here.
@pcc since you reviewed the above-mentioned patch - could you please take a look at this one too?
Oct 17 2019
I think you could've just used CHECK-DAG to fix the tests. It *might* be a bit more robust. Although just reordering checks seems perfectly fine too.
Oct 16 2019
Looks great! Thanks!