David Daney wrote:
Joe Buck wrote:
Maybe there could be a "semi-primary" or "experimental primary" status;
a feature could be treated as primary, but with the understanding that
the requirement will be waived if it causes excessive delay. The
"experimental" label could be dropped after a few successful releases.
>
What is the real purpose of being a primary language?
From my point of view essentially only that regressions can be marked
P1 to P3 and are thus on the radar of release manager. If some serious
wrong-code regression has been found, I think it is good if it shows up
as serious regression and not simply gets the comment
"Fortran is not release critical -> P4".
If Fortran were release critical, bugs could still be downgraded to P4 -
either because it is seen as no release critical bug or no patch could
be provided in time. (I think that Fortran PRs are a bit more likely to
be downgraded by the RM than C PRs thus it would be anyway effectively a
semi-primary language. Besides, also for primary languages the number of
regressions is not zero and there are test-suite failures.)
So far, fortunately there were only few "release-critical" Fortran
regressions and there were fixed quickly thus there was no problem in
practice.
Regarding the number of test-suite failures: I think the number of
failures is relatively low (see François-Xavier's email); we try to
reduce them to zero, but without direct access to failing systems it
takes a bit longer. Also several failures are libc problems, though one
needs to verify this - on one system it might be due to a libc bug but
on another it is a real bug (which is on the first system either not
present or hidden by the libc bug).
> The only benefit I can really see is that it makes good PR.
For me this is less important, but granted it is also a benefit (-:
Tobias