Milder CVS problems, redux

Milder CVS problems, redux

[JD Smith]

I had previously reported problems with CVS commands in Emacs 22
timing out, and/or the cvs process getting killed.  They didn't recur,
but I *do* now commonly get empty results from the first run of cvs
log (C-x v l) on a local repository, which causes a two line empty
buffer to display.  Re-running the command usually succeeds, but it
seems very likely to me that these problems may be related, perhaps to
some underlying race condition.  

I see that CVS log is running asynchronously (vc-cvs.el):

  (vc-cvs-command
   buffer
   (if (and (vc-stay-local-p file) (fboundp 'start-process)) 'async 0)
   file "log"))

Explicitly removing 'async "fixes" this problem, but I presume log is
async for a reason, and that this shouldn't introduce race conditions
of this sort.  I see that diff can operate asynchronously as well, but
that others do not, which is consistent with my earlier findings
("log" and "diff" were the problem cases).  Unfortunately I don't have
much experience debugging asynchronous processes.  I can consistently
reproduce the "blank buffer" issue on C-x v l as of now.

Re: Milder CVS problems, redux

[Chong Yidong]

JD Smith <jdsmith@as.arizona.edu> writes:

> I had previously reported problems with CVS commands in Emacs 22
> timing out, and/or the cvs process getting killed.  They didn't recur,
> but I *do* now commonly get empty results from the first run of cvs
> log (C-x v l) on a local repository, which causes a two line empty
> buffer to display.  Re-running the command usually succeeds, but it
> seems very likely to me that these problems may be related, perhaps to
> some underlying race condition.  
>
> I see that CVS log is running asynchronously (vc-cvs.el):
>
>   (vc-cvs-command
>    buffer
>    (if (and (vc-stay-local-p file) (fboundp 'start-process)) 'async 0)
>    file "log"))
>
> Explicitly removing 'async "fixes" this problem, but I presume log is
> async for a reason, and that this shouldn't introduce race conditions
> of this sort.  I see that diff can operate asynchronously as well, but
> that others do not, which is consistent with my earlier findings
> ("log" and "diff" were the problem cases).  Unfortunately I don't have
> much experience debugging asynchronous processes.  I can consistently
> reproduce the "blank buffer" issue on C-x v l as of now.

Are you sure you have the latest version of Emacs from CVS?  This
sounds like a bug I fixed last week; see

  http://lists.gnu.org/archive/html/emacs-devel/2006-07/msg00588.html

Re: Milder CVS problems, redux

[JD Smith]

On Mon, 17 Jul 2006 17:49:01 -0400, Chong Yidong wrote:

> Are you sure you have the latest version of Emacs from CVS?  This sounds
> like a bug I fixed last week; see
> 
>   http://lists.gnu.org/archive/html/emacs-devel/2006-07/msg00588.html

Right you are, this fixes it for me.  I am also cautiously hopeful your
patch addresses the whole category of problems relating to cvs log getting
out of sync.

Thanks,

JD