[llvm-commits] [llvm] r163098 - /llvm/trunk/test/Makefile

[Rick Foos]

On 09/06/2012 10:41 AM, Will Schmidt wrote:
> On Thu, 2012-09-06 at 10:16 -0500, Rick Foos wrote:
>> On 09/05/2012 04:16 PM, Will Schmidt wrote:
>>> On Wed, 2012-09-05 at 15:26 -0500, Rick Foos wrote:
>>>> So switching from a double quote within a double quoted string to a
>>>> single quote avoided that problem.
>>>>
>>>> Shell interferes with escapes, and thus the 5 \\\\\'s to try to do
>>>> what
>>>> is supposed to work in 3,  '\\\"'. The extra subst wrapper avoids the
>>>> problem, and results in something that has a better chance of working
>>>> everywhere.
>>>>
>>>> So your buildslave is fine. It would be good to run the change on
>>>> your
>>>> machine to get more proof that this change works everywhere.
>>>>
>>>> Cheers,
>>>> -rick
>>>>
>>> Hi Rick,
>>>     Not sure I've followed your description and/or tested the subst()s
>>> quite right.   I came up with:
>>> +	@$(ECHOPATH) s=@OCAMLOPT@=$(OCAMLOPT) -cc "$(subst *,'\\\"',*$(subst =,"\\\=",$(CXX_FOR_OCAMLOPT))*)" -I $(LibDir)/ocaml=g>>   lit.tmp
>>>
>>> and get lit.tmp entry as:
>>> s=@OCAMLOPT@=/bin/ocamlopt -cc '\"'g++ -Wall -D_FILE_OFFSET_BITS\=64 -D_REENTRANT -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE\=2 -fexceptions -fstack-protector --param\=ssp-buffer-size\=4  -m64'\"' -I /home/willschm/llvm/Release+Asserts/lib/ocaml=g
>>>
>>> and lit.site.cfg entry as:
>>> config.ocamlopt_executable = "/bin/ocamlopt -cc '"'g++ -Wall -D_FILE_OFFSET_BITS=64 -D_REENTRANT -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4  -m64'"' -I /home/willschm/llvm/Release+Asserts/lib/ocaml"
>>>
>>> which seems good enough to run the tests, but doesn't quite match the
>>> escaped pattern you showed in your email:
>>> config.ocamlopt_executable = "/usr/bin/ocamlopt -cc \"g++
>>> -D_FILE_OFFSET_BITS=64 -D_REENTRANT\" -I
>>> /local/jenkins-swarm-u1204/workspace/llvm-repo/build-x86_64-linux-gnu/Release+Asserts/lib/ocaml"
>>>
>>> If you can share your exact diff, i'll give that a spin here and verify
>>> its OK.
>>>
>>> Thanks,
>>> -Will
>>>
>>>
>>>
>> Were you able to run with the patch, or have an alternative?
> The change I came up with was short one "\" with respect to your patch,
> which probably (hopefully) explains why my testrun yesterday didn't look
> right.
>
>
> With your change applied:
>
> lit.tmp:
> s=@OCAMLOPT@=/bin/ocamlopt -cc \\\"g++ -Wall -D_FILE_OFFSET_BITS\=64
> -D_REENTRANT -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE\=2 -fexceptions
> -fstack-protector --param\=ssp-buffer-size\=4 -m64\\\"
> -I /home/willschm/llvm/Release+Asserts/lib/ocaml=g
>
> lit.site.cfg:
> config.ocamlopt_executable = "/bin/ocamlopt -cc \"g++ -Wall
> -D_FILE_OFFSET_BITS=64 -D_REENTRANT -O2 -g -pipe -Wall
> -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector
> --param=ssp-buffer-size=4 -m64\" -I /home/willschm/llvm/Release
> +Asserts/lib/ocaml"
>
> Does the lit.site.cfg line look OK to you?
>
>   - we fell off the -commits list on my earlier reply to you..  go ahead
> and CC -commits on your reply if/as you wish.  :-)
>
> Thanks,
> -Will
>
Hi Will,

Yes, I dropped off the list until we confirmed things. If something 
wasn't right it could take a while to duplicate and fix.

Your results look good now, and I think it's OK to apply the patch as-is!

Can you help apply this Nuno?

Thanks,
-Rick

-- 
Rick Foos
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by The Linux Foundation