[PATCH 5 of 5 V3] tests: check that procutil.std{out,err}.write() returns correct result

[Yuya Nishihara]

On Wed, 15 Jul 2020 14:23:17 +0200, Manuel Jacob wrote:
> On 2020-07-15 13:14, Yuya Nishihara wrote:
> > On Tue, 14 Jul 2020 20:41:51 +0200, Manuel Jacob wrote:
> >> # HG changeset patch
> >> # User Manuel Jacob <me at manueljacob.de>
> >> # Date 1594291924 -7200
> >> #      Thu Jul 09 12:52:04 2020 +0200
> >> # Node ID b2aabfc8d6800bf6e103b783a245c45e3b1ae4ac
> >> # Parent  49ebcf3b048c7352a4c32b1fd6c793ab99aec246
> >> # EXP-Topic stdio
> >> tests: check that procutil.std{out,err}.write() returns correct result
> > 
> >> -procutil.{stream}.write(b'x' * 1048576)
> >> +write_result = procutil.{stream}.write(b'x' * 1048576)
> >> +with open({write_result_fn}, 'w') as write_result_f:
> >> +    write_result_f.write(str(write_result))
> > 
> > Nit: {write_result_fn!r} can be used instead of repr().
> > 
> >> +            # tempfile.mktemp() is unsafe in general, as a malicious 
> >> process
> >> +            # could create the file before we do. But in tests, we're 
> >> running
> >> +            # in a controlled environment.
> >> +            write_result_fn = tempfile.mktemp()
> > 
> > Even so, I think it's better to not use deprecated function which may 
> > be
> > deleted in future version.
> 
> According to the module documentation, the file created by the more 
> secure (for general use) variants can’t be opened twice, but that’s 
> exactly the functionality we need. For this reason, I don’t think 
> tempfile.mktemp() will be deleted. But I can try to find some hack.

IIUC, we can just create an empty temp file (and close it immediately)
to reserve the file name.