[LLVMdev] [llvm-commits] [PATCH] A "very verbose" mode for FileCheck

Dmitri Gribenko gribozavr at gmail.com
Wed Jan 16 16:02:36 PST 2013

On Thu, Jan 17, 2013 at 1:55 AM, Evgeniy Stepanov
<eugeni.stepanov at gmail.com> wrote:
> On Thu, Jan 17, 2013 at 3:31 AM, Dmitri Gribenko <gribozavr at gmail.com> wrote:
>> On Thu, Jan 17, 2013 at 1:23 AM, Evgeniy Stepanov
>> <eugeni.stepanov at gmail.com> wrote:
>>> On Thu, Jan 17, 2013 at 1:19 AM, Dmitri Gribenko <gribozavr at gmail.com> wrote:
>>>> On Wed, Jan 16, 2013 at 9:33 PM, Dmitri Gribenko <gribozavr at gmail.com> wrote:
>>>>> On Wed, Jan 16, 2013 at 9:24 PM, Chris Lattner <clattner at apple.com> wrote:
>>>>>> On Jan 16, 2013, at 10:32 AM, Dmitri Gribenko <gribozavr at gmail.com> wrote:
>>>>>>> Hello,
>>>>>>> When someone breaks a FileCheck-based test on some buildbot, sometimes
>>>>>>> it may not be obvious *why* did it fail.  If the failure can not be
>>>>>>> reproduced locally, it can be very hard to fix.
>>>>>>> I propose adding a "very verbose" mode to FileCheck.  In this mode
>>>>>>> FileCheck will dump the input file in case of failure.  This mode will
>>>>>>> be enabled by an environment variable "FILECHECK_VERY_VERBOSE".  If we
>>>>>>> chose a command line option, we would have to edit all FileCheck-based
>>>>>>> tests to use %FileCheck.
>>>>>> I think that this idea is good, but I'd prefer it be implemented a different way:
>>>>>>  - Filecheck should take a new flag -dump-input-on-error that causes it to... dump the input file on error.
>>>>>>  - Lit should be the thing that checks the environment (or perhaps add a new option to lit), and adds the flag to FileCheck invocations.
>>>>>> I don't like it when the behavior of such a low-level tool like this changes based on environment variables.  It isn't discoverable in --help.  If for some reason, it is bad for lit to implicitly pass the option, I'd rather have a standard FILECHECK_COMMANDLINE environment variable, and have filecheck parse arbitrary options out of it using the cl::ParseEnvironmentOptions function.
>>>>> I agree that a command line option would be better.  But in that case
>>>>> all tests should be updated.  It is not an issue for me -- it is
>>>>> mostly mechanical.  So should I change tests to use %FileCheck?
>>>> Here's a third attempt.
>>>> The new behavior is as follows:
>>>> 1. In case of errors we always dump output to a temporary file and print
>>> Does it mean we get one more file in /tmp every time a test fails, and
>>> it is not cleaned up automatically? I don't think this should happen
>>> in the "default" mode of the tool.
>> Well, yes.  David requested that and I agreed that it is a good idea.
>> Are you strongly opposed to it?
> Not really, I doubt it can break something for someone...
> But if we only need this behaviour on the buildbot, and you are going
> to update all tests anyway, why not make it conditional on a flag?

Buildbots don't need to create files.  Richard requested this because
it is just generally useful -- usually the first thing you do after
seeing a failed FileCheck test is running the command manually to see
the output.


(j){printf("%d\n",i);}}} /*Dmitri Gribenko <gribozavr at gmail.com>*/

More information about the llvm-dev mailing list