Re: pkg/25317

[Alan Barrett <apb%cequrux.com@localhost>]

On Tue, 29 Mar 2005, kim%netbsd.org@localhost wrote:
> Synopsis: spurious "No more PTYs" from screen
> 
> It is a bit difficult to understand the patch to a patch, i.e. I'm
> having trouble figuring out which method you want to use.

To properly see what I was talking about, you need to check out a
version of pkgsrc from the date when I filed the PR.  Then apply the
patch-to-a-patch, examine the resulting patch file to see how it would
patch the original sources, apply the new patch file to the correct
version of the original sources, examine the reulting patched source
code.  And compare that with the resulting patched source code that you
would get without using the patch-to-a-patch.

> Looking at pty.c the only branch for "#ifndef PTY_DONE" is the last
> one that defaults to BSD style PTY's.  It doesn't use openpty(3),
> so I don't understand why you'd change that to require HAVE_OPENPTY
> defined.

Huh?  I didn't change that block at all.  I think you are looking at a
version of pty.c as patched by revision 1.10 of patch-aa, whereas my PR
was against revision 1.9 of patch-aa.

> The NetBSD block was added above so that the default case would not
> be used on NetBSD, as it doesn't understand PTY naming on NetBSD.

The bug which I reported is that the "#if defined(NetBSD)" block was
*below* the "#if defined(HAVE_OPENPTY)", whereas it should have been
*above* the "#if defined(HAVE_OPENPTY)".  You say the NetBSD block
was added above, I say that it *should have* been added above, and it
probably was above in an older version of screen, but (with the versions
of screen and patch-aa against which I filed the PR), the NetBSD block
ended up being added below instead of above the HAVE_OPENPTY block.  The
whole point of the PR was to report that problem.

The reason that the patch used to apply in the wrong place is that the
pty.c file contains many identical lines.  For example, the three-line
sequence {blank line, comment line with many stars, blank line} appears
many times.

I believe that the current version of patch-aa will not apply in
the wrong place, so the PR can be closed.  However, it's really
disappointing that PRs that contain patches just get ignored for such a
long time.  The longer my PRs get ignored, the more difficult it becomes
to answer questions about them, and the more inclined I am to keep my
patches to myself instead of submitting PRs.

> I have a feeling that the "other" openpty(3) block got added in a
> later version of screen(1), so most likely the NetBSD block is no
> longer used.  However, I can't test this without an old enough
> system, e.g. NetBSD 1.6.2.

Yes, I think that's true.  I also no longer have a sufficiently
old version of NetBSD.

--apb (Alan Barrett)