[cfe-commits] [PATCH] expected-no-diagnostics for -verify (was: Re: r164677 - )

[David Blaikie]

looks fairly reasonable, though I'd expect Doug or someone to sign off
on a new feature(tte) like this

Some nits/thoughts:
* VerifyDiagnosticConsumer.cpp: braced one line else (usually we drop
the braces from one line blocks) around "Status =
...HasExpectedNoDiagnostics"
* verify3.c:
  * using -D to separate multiple tests within a single file can be a
bit hard to follow. Would this be possible/easier to read if they were
just separate files?
  * Also the right-alignment of CHECK prefixes isn't common, though
not inherently problematic. (just seems like it'd make for some
maintenance difficulty in the future should any check prefix become
longer than the longest one already in use)
* It might be nice to commit the expected-no-diagnostics fixes first,
then commit the improvement just to separate the mechanical from the
interesting work.
* verify.m: is there a reason you didn't use expected-no-diagnostics there?

On Wed, Oct 10, 2012 at 1:21 AM, Andy Gibbs <andyg1001 at hotmail.co.uk> wrote:
> ** bump ** :o)
>
> I've had to resync and attach the patch again here; a test-case needed to be
> updated to reflect changes in trunk.  The patch now is taken against
> r165609.
>
> Cheers
> Andy
>
>
> On Saturday, October 06, 2012 10:59 PM, Andy Gibbs wrote:
>>
>> Hi,
>>
>> Was there any interest in this patch?  Please let me know if there are any
>> changes that should be made, or whether it is good to be committed? (or even
>> whether I was barking up the wrong tree...?!)
>>
>> (If any were unable to receive the patch -- I know that sometimes Apple
>> users have problems with my attachments! -- then it can be viewed directly
>> at [...snip...])
>>
>> Cheers
>> Andy
>>
>>
>> On Tuesday, October 02, 2012 5:32 PM, Andy Gibbs wrote:
>>>
>>> All,
>>>
>>> Attached is a patch that adds an 'expected-no-diagnostics' directive
>>> to -verify.  This can then be used, as discussed, to flag files that are
>>> not
>>> expected to produce any diagnostics at all.  -verify has then been
>>> changed
>>> to generate an error if no expected-* directives are found (for example,
>>> if
>>> the RUN line doesn't specify the source file!).  I've also made it so
>>> that
>>> the user cannot put 'expected-no-diagnostics' in with other expected-*
>>> directives (to avoid mistakes!).
>>>
>>> The complete patch is rather large, but I've grouped the changes to
>>> VerifyDiagnosticConsumer to the top of the file along with the additional
>>> test-case.  The rest of the patch makes the necessary changes to fix 580
>>> test-cases which now (in almost all cases) need to have the
>>> 'expected-no-diagnostics' directive.
>>>
>>> Comments / questions welcome!
>>>
>>> Andy