[PATCH] [libcxx] Use pseudo-terminal in LIT tests so color output from compiler error messages is captured.

[David Blaikie]

On Mon, Oct 27, 2014 at 6:51 PM, Eric Fiselier <eric at efcs.ca> wrote:

> >But what if Lit isn't running in a device capable of color output? Won't
> >this add a bunch of ugly ASCII codes?
>
> Wouldn't clang refrain from outputting color codes to begin with?
>

I believe it doesn't know any better, if you've given it a pseudoterminal
capable of colours, it'll go for it.


> Do you know of any way to detect this?
>

The same way clang, lit, and FileCheck detect it, I suppose... (I'm sure
there's some API in libSupport, though I realize that won't be directly
usable from this python, but might give you some inspiration)


>
> >I think what tools like Ninja do is to also optionally strip the color
> >codes in that case (
> >https://github.com/martine/ninja/blob/master/src/util.cc#L384).
>
> Not much you can do about that. At least the colored output makes it one
> step further.
>

Agreed (& note that FileCheck is also a great candidate for this). For now
I just use the compiler flags in cmake to force Clang's colours on. This
doesn't work if I want to redirect, of course, and doesn't work for
FileCheck (due to lit) nor lit (due to Ninja) - having an option to force
these things on would be fine by me too, so I could do the same with those
tools as I do with clang (& indeed if you make this change to lit and give
me a "force colours" flag (or manage to get it in without the terminal
detection so it's just always on - but I suspect that'd be bad, as Daniel
pointed out) then I might not need a flag for FileCheck and I can remove
the flag passing I have for clang... )


>
> Thanks for the input.
>
> http://reviews.llvm.org/D6010
>
>
>
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20141027/8f9fc17b/attachment.html>