This is the mail archive of the fortran@gcc.gnu.org mailing list for the GNU Fortran project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Proposed renaming of testsuite files


2009/7/19 Jerry DeLisle <jvdelisle@verizon.net>:
> I noticed a comment or two some time back about naming test cases after the
> PR number and/or otherwise re-organizing.
>
> I thought I would suggest an approach.
>
> 1) Many of the test cases currently contain a reference to there associated
> PR number. ?I suggest we automate a script to extract that PR number out,
> rename the file as that number, and store the old file name as an index in
> the file.

I think most comments that I heard were actually *against* naming test
cases after PR numbers, and IMHO it is indeed preferable to have
descriptive names instead of pure numbers. If you have lots of test
cases this makes it much easier to distinguish them.

And in case you look for the test case corresponding to a certain PR,
you can still grep for the number (I always put the PR number in a
comment inside the test case) or just look at the commit message.


> 2) Define a set of "categories" for test cases and sort them into a few
> subdirectories. ?Tow shorten editor load time when opening files.

One possible scheme would be to sort test cases by the version of the
Fortran standard which is required to compile them. We could e.g. have
subdirectories for f77/f95/f03/f08. In a way this would be in line
with our current naming scheme, which thematically groups test cases
containing the same Fortran 'features'.

This also would make it easier to make statements like: "Currently
gfortran is in perfect shape as a F95 compiler, but we have a few
failures in the F03 sector." If someone has little time for running
the whole testsuite and is only interested in gfortran as a F95
compiler, he might only run the corresponding parts of the testsuite
(i.e. f77 and f95). Also then one could argue that F95 regressions are
more severe than F08 regressions etc (because F95 is more widely used
...)

Of course this scheme might also have drawbacks, but it is the one
which would make the most sense to me.

Cheers,
Janus


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]