Rolling back the thread a bit...

On Wednesday, July 04, 2012 2:46 AM, Jordan Rose wrote:
> On Jul 3, 2012, at 4:08 PM, Andy Gibbs wrote:
>> I think a problem I have seen is that where a fatal error occurs, you
>> simply get no diagnostics at all -- except a cryptic line telling you how
>> many errors occurred.  From my point of view it is much better to always
>> see the -verify diagnostics since this is the purpose of the feature --
>> to catch diagnostics and display them where they do not match those
>> expected.  In the current case, you don't even get the diagnostic of the
>> line *causing* the fatal error!
> 
> Again, what I'd prefer is that fatal errors are NOT suppressed, or rather
> not caught by VerifyDiagnosticsConsumer, so even if you don't get the
> verify-related complaints you still know what's up. Something as simple as
> a check in HandleDiagnostic for DiagLevel == FatalError, and then a very
> simple text output on the spot would do it for me.
> 
> However, if we don't go with that, this approach is better than the last
> one (though I would have personally used a flag argument on
> EmitCurrentDiagnostic rather than a separate ForceEmitCurrentDiagnostic).
> I can live with it. (Anyone else have opinions here?)

Ok, attached is the patch referred to in the above conversation, but with
the fix suggested implemented.

Cheers
Andy

Attachment: verify-part4.diff
Description: Binary data

_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to