[compiler-rt] r264773 - [profile] Make a test work if run by the super-user

[Vedant Kumar via llvm-commits]

> On Mar 29, 2016, at 1:10 PM, Sean Silva <chisophugis at gmail.com> wrote:
> 
> 
> 
> On Tue, Mar 29, 2016 at 1:03 PM, Vedant Kumar <vsk at apple.com> wrote:
> 
> > On Mar 29, 2016, at 1:00 PM, Sean Silva <chisophugis at gmail.com> wrote:
> >
> >
> >
> > On Tue, Mar 29, 2016 at 12:55 PM, Vedant Kumar <vsk at apple.com> wrote:
> > Darwin is funny. I tried an empty string also, but the test failed :(.
> >
> > Interesting. Are the quotes necessary? For the record, for PS4, both `"/"` (your patch), `/`, and `` (i.e. empty) work.
> 
> It actually fails (i.e, proceeds with the write) with and without the quotes.
> 
> Sorry, maybe I should clarify. For my testing on PS4, this test works fine with either `"/"` (your patch), `/`, and `` (i.e. empty). Did you intend to say that `"/"` works (i.e. the test case passes) on Darwin? I'm not sure what to make of your statement "with and without the quotes".

Yes, this test case passes on Darwin (i.e, with `"/"`). However, it does not pass with `` or `""` on Darwin.


> > That said, it's important the test is portable. I'll revert and try again if the Windows bots have issues.
> >
> > I haven't tested a windows-targeting configuration (and don't know if we have a bot for that). The closest I have is windows-hosted targeting PS4.
> 
> Hm, it could be that the public Windows bots have never run check-profile. The old version of the code used chmod etc. without a `REQUIRES: shell` line.
> 
> Should we add a require line to avoid breaking future Windows bots?
> 
> Let's see if anything breaks first.

Ok, fair enough.

vedant