Re: [2.20] [4/6] Enumerate tests with special rules in tests-special variable

["Joseph S. Myers" <joseph at codesourcery dot com>]

On Thu, 6 Mar 2014, Carlos O'Donell wrote:

> > * Should tests that don't generate output files be changed (in a
> >   separate patch or patches) to do so?
> 
> Yes. All tests should generate output files. I also think all tests
> should be named in such a way that makes it easy to identify the test
> e.g. starting with tst-, and not bug-. As it makes it easier for
> developers to know exact what is and is not a test, grep, sed, and
> do other manipulations on them for whatever purposes.

If you want a convention for naming source files, I wonder about using a 
test/ subdirectory of each directory, for all files used only for tests 
(including putting sysdeps files used only for tests in test/ 
subdirectories of sysdeps directories - though this would make the paths 
to abilist baselines even longer), and likewise in object directories.

> > +# Run a test on the header files we use.
> > +# XXX Please note that for now we ignore the result of this test.
> 
> Nit. Please make this a TODO: marker and describe briefly that we need
> to check the result of this test, which we presently ignore.

The patch is moving this comment from one place to another, not changing 
it (and in any case what we actually need to do is obsolete this test by 
conformtest, which I hope to do eventually after getting a copy of 
POSIX.1-1990 which defines _POSIX_SOURCE as used by this test - 
unfortunately the copy I ordered in January hasn't turned up yet).

-- 
Joseph S. Myers
joseph@codesourcery.com