[llvm-commits] patch to enable response file support in ParseCommandLineOptions

[Duncan Sands]

Hi Sam,

On 02/10/12 17:18, Liu, Yaxun (Sam) wrote:
> Thanks Chris for the comment.
>
> Since there is no objection, I attached a new patch which enables response file
> support and removes the argument for controlling/disabling response file
> support. The patch also contains a simple test. I did regression check and there
> are no regressions.

the patch looks good but I don't think the test will work everywhere:

+; Test response file with llvm-link
+; RUN: i=1; while [[ $i -lt 10 ]]; do echo "define void @foo$i(){ret 
void}"|llvm-as -o %t.foo$i.bc; i=$((i+1)); done
+; RUN: rm -f %t.bar.ll
...

You can't rely on the RUN lines being executed by a sensible shell, thus
anything complicated, eg these loops, isn't going to work universally.
How about expanding out the loop by hand into many more RUN lines instead.

Ciao, Duncan.

>
> + llvmdev
>
> Sam
>
> *From:*Chris Lattner [mailto:clattner at apple.com]
> *Sent:* Sunday, September 30, 2012 2:20 PM
> *To:* Liu, Yaxun (Sam)
> *Cc:* llvm-commits at cs.uiuc.edu
> *Subject:* Re: [llvm-commits] patch to enable response file support in
> ParseCommandLineOptions
>
> On Sep 21, 2012, at 2:07 PM, "Liu, Yaxun (Sam)" <Yaxun.Liu at amd.com
> <mailto:Yaxun.Liu at amd.com>> wrote:
>
>
>
> Hi,
>
> I am sending a patch to enable response file support in ParseCommandLineOptions.
> With this change, all llvm tools will support response file. This helps overcome
> the command line length limit which we encountered recently.
>
> This seems fine to me.  Is there any reason we *wouldn't* want a tool to read
> response files?  If all tools should handle it, the argument should just go away.
>
> -Chris
>
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>