[LLVMdev] Win32 COFF Support

Nathan Jeffords blunted2night at gmail.com
Mon Jun 14 13:38:25 PDT 2010


>
>
> > For the debug macros, I personally like them because they reduce the
> impact debug code has on the "real" code. I find dbgout("message") simpler
> and more concise than DEBUG(dbgs() << "message") for the consumer of the
> debug API. The other macros are there to not have to reproduce the same
> pattern a repeatedly.
>
> Are these intended to stay permanently or are they just for bring-up?  Your
> argument implies that we should pull them into a higher-level header file,
> but personally I think that explicitly using "dbgs() <<" makes it clearer
> what is happening.
>
>
I don't think there would be alot of confusion about the purpose of dbgout.
In my projects, dbgout is always globally available. Its implementation
changes based on what debug facilities are available, but its interface
remains the same. I would like to see this put into a higher level header
file.

Now that someone has provided a wrapper MCStreamer that dumps the calls
being made, the other macros are not as important as they mostly just dump
the incoming parameters. Though having to have all the calls that didn't
fail dump their data just to see that data of the one that did fail might
not be ideal.

dbgout_calls can definitely go away. I think the other two (dbgerr,
dbg_notimpl) are useful at least locally and would like to keep them in some
form.

-Nathan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20100614/9aaf6ed2/attachment.html>


More information about the llvm-dev mailing list