[llvm-lit] remove %t files before running a test

[James Molloy]

Hi,

OK, your original patch was clearer. I'd prefer to go with that.

Cheers,

James

On Thu, 14 May 2015 at 15:34 Tzafrir Poupko <tzafrir11 at gmail.com> wrote:

> Hi James,
>
> As far as I can tell rmtree() does not support removing symbolic links or
> matching files by pattern.
> Since i would like to support removal of patterns such as %t , %t.txt and
> %t/filename I don't see how rmtree is useful by itself
>
> Attached another patch option implementing the file scanning in python
> instead of calling rm -rf
>
> On Thu, May 14, 2015 at 4:00 PM, James Molloy <james at jamesmolloy.co.uk>
> wrote:
>
>> Hi Tzafrir,
>>
>> In which case, just using shutil.rmtree() should be sufficient:
>> https://docs.python.org/2/library/shutil.html#shutil.rmtree
>>
>> Cheers,
>>
>> James
>>
>> On Thu, 14 May 2015 at 13:57 Tzafrir Poupko <tzafrir11 at gmail.com> wrote:
>>
>>> Hi James,
>>>
>>> There are 2 reasons for the -f
>>>
>>> 1. This is what is done currently by more then 300 tests and I wouldn't
>>> like to change this.
>>> 2. Using -f suppresses error messages that will happen when no temporary
>>> file exists
>>>
>>> Since there is no standard function in python to delete files and
>>> directories matching a pattern
>>> I thought using rm -rf would be a better solution then implementing my
>>> own.
>>>
>>> Tzafrir
>>>
>>> On Thu, May 14, 2015 at 3:27 PM, James Molloy <james at jamesmolloy.co.uk>
>>> wrote:
>>>
>>>> Hi Tzafrir,
>>>>
>>>> This sounds like a good idea to me.
>>>>
>>>> +    subprocess.call(['rm','-rf',path+"*"])
>>>>
>>>>
>>>>
>>>> This is super dangerous. Why the -f ?
>>>>
>>>> Cheers,
>>>>
>>>> James
>>>>
>>>> On Thu, 14 May 2015 at 13:22 Tzafrir Poupko <tzafrir11 at gmail.com>
>>>> wrote:
>>>>
>>>>> Before running a test, purge all %t files used by previous tests.
>>>>> Remove any custom file used in test which was prefixed by %t.
>>>>> There are over 300 clang tests that run "rm -rf %t" as first run
>>>>> command, this patch will make it automaticlly.
>>>>>
>>>>> _______________________________________________
>>>>> llvm-commits mailing list
>>>>> llvm-commits at cs.uiuc.edu
>>>>> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>>>>>
>>>>
>>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150514/e7c94f64/attachment.html>