[PATCH][libcxx] scan-build fixes in tests

[Eric Fiselier]

Hi Steve,

It seems most of the changes just remove unused variables after they are
initialized. Isn't the initialization part of the test?
For example, if the return type of `foo()` is specified to be a type that
is BooleanConvertible then doesn't it make sense to test `bool unused =
foo()`?
I would rather just see the variables be used as opposed to their ommision.
Even if using them just means `((void)unused)`.

/Eric

On Fri, Oct 17, 2014 at 7:24 PM, Steve MacKenzie <stevemac321 at live.com>
wrote:

> I am about half-way through running the libc++ tests with scan-build.  I
> thought it good to submit a patch now for review for feedback rather than
> later.
>
> All the bugs are of the "dead initialization" category. I tried to apply
> the least intrusive fix as possible, which in all but a couple cases,
> involved simply removing the unused lvalue from the statement.  So the
> coverage should remain the same, neither augmented nor diminished.
>
> I did encounter one scan-build false positive, it looks to have been
> reported already (10862). I added a very simple repro to the bug report. (I
> thought about adding a comment in these tests referencing the bug 10862,
> but did not, let me know if I should).
>
> The status of the run is being tracked on my blog here:
> http://stevemac123.wordpress.com/static-analysis-run/
>
> clang version 3.6.0 (trunk 217475)
> Target: x86_64-unknown-linux-gnu
> Thread model: posix
>
> Command line used:
>
> can-build -k -V -analyze-headers clang++ -std=c++1y -stdlib=libc++
>
> Thanks,
> Steve MacKenzie
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20141017/19cca546/attachment.html>