llvm dev - May 2008 - [LLVMdev] DejaGnu test-suite coverage

Heh,

my ex-boss would die seeing something like this :-)

Just one comment, does GCOV have a flag for
a third category of lines, like "undesired to execute"?

This would make the summary of
<http://chandlerc.net/llvm-coverage/lib/Target/PowerPC/ 
PPCHazardRecognizers.cpp.gcov.html>
much more favorable.

Nice work!

Cheers,

	Gabor

On Mon, May 26, 2008 at 11:27 PM, Gabor Greif <gabor at mac.com> wrote:
> Heh,
>
> my ex-boss would die seeing something like this :-)

It's not too hard to produce, let me know if you want details, or ping me on
IRC.

> Just one comment, does GCOV have a flag for
> a third category of lines, like "undesired to execute"?

Nope. I'm a subscriber to the theory that everything should be executed,
right down to assertions, and compile assertions. ;] Not always possible,
but that is usually my ideal goal in testing. The key is to always check
that the result of that execution is correct, so trigger assertions, and
catch them, etc.

> This would make the summary of
> <http://chandlerc.net/llvm-coverage/lib/Target/PowerPC/
> PPCHazardRecognizers.cpp.gcov.html>
> much more favorable.

I'm not seeing the bad aspect of this particular file? It has pretty good
coverage, is code calling into this "bad"? If so, then assertions, or
logging or something might be more appropriate. Coverage just says that it
got used, not that its use was "correct". ;]

-Chandler

> Nice work!
>
> Cheers,
>
>        Gabor
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20080526/8bfb9eb1/attachment.html>

On May 26, 2008, at 11:55 PM, Chandler Carruth wrote:
> This would make the summary of
> <http://chandlerc.net/llvm-coverage/lib/Target/PowerPC/
> PPCHazardRecognizers.cpp.gcov.html>
> much more favorable.
>
> I'm not seeing the bad aspect of this particular file? It has pretty  
> good coverage, is code calling into this "bad"? If so, then  
> assertions, or logging or something might be more appropriate.  
> Coverage just says that it got used, not that its use was  
> "correct". ;]
I think he means the lines like:

switch (Opcode) {
default: assert(0 && "Unknown load!");

where the default case is never executed.
-Chris
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20080527/96d6be36/attachment.html>