Re: I cannot run zsh: child state waiting for longjmp

["Peter A. Castro" <doctor@xxxxxxxxxxxx>]

On Thu, 5 Aug 2010, Peter Stephenson wrote:

> (Moved to zsh-workers.)
>
> On Wed, 4 Aug 2010 15:12:12 -0700 (PDT)
> "Peter A. Castro" <doctor@xxxxxxxxxxxx> wrote:
> > On Mon, 2 Aug 2010, Peter Stephenson wrote:
> > > > I built zsh-4.3.10 on MSYS on Win XP.
> > > > It's working fine but Whenever I set path to whatever value, let's
> > > > say: path=(. /usr/local/bin /mingw/bin /bin)
> > > >
> > > > and run whatever command, let's say ls, I get:
> > > >
> > > > --------------------
> > > > $ path=(. /usr/local/bin /mingw/bin /bin)
> > > > $ ls
> > > >        0 [main] zsh 1556 sync_with_child: child 444(0x6AC) died
> > > > before initialization with status code 0xC0000135
> > > >    17380 [main] zsh 1556 sync_with_child: *** child state waiting
> > > > for longjmp
> > > > zsh: fork failed: resource temporarily unavailable
> > > > --------------------
> >
> > Unfortunately, I'm seeing this same kind of symptom under Cygwin
> > (1.7.5) when I run 'make check'.  Several of the tests generate
> > something similar to the above while others just plain hang and I
> > have to kill the subshells in order for it to continue.  I'm not sure
> > how much of this is Cygwin and how much of it is zsh.  I'll attempt
> > to debug it but I don't have the time right now.
>
> I'm not seeing any problems this basic with the latest zsh code on my setup
> with Windows XP and a Cygwin from April (setup.exe claims it's 1.7.5-1).  I

Hmm... I'll see if I can rustle up some output to show it (machine is
down at the moment).  Cygwin 1.7.5-1 is the current runtime, but I've
seen these issues as early as 1.7.0 (which is why I hadn't released it
4.3.10 for Cygwin until very recently).

> am seeing failing tests with process substitutions hanging; we've had this

Yes, I've seen this too.  I remember seeing some discussion about why
this might be hanging but thought it might have some relationship to the
I/O redirection issues I've been seeing.  Just a guess, really.

> issue before and been back and forth between the two possible
> implementations to try to fix it.  Apart from that, I don't think we're
> going to get further within zsh without more detail of the failures:
> relevant Cygwin changes / Windows versions / zsh configurations.

I agree.  I'll see if I can get some debug time put in to it.  :-)

> Here's a change for the MACHINES file.  I think the paragraph I've removed
> is well out of date.

Ah, so, we cover it up in doc and claim it as a "feature"?  Good, good :-)

> Index: MACHINES
> ===================================================================
> RCS file: /cvsroot/zsh/zsh/MACHINES,v
> retrieving revision 1.8
> diff -p -u -r1.8 MACHINES
> --- MACHINES	16 May 2009 12:13:00 -0000	1.8
> +++ MACHINES	5 Aug 2010 12:02:31 -0000
> @@ -46,10 +46,13 @@ Red Hat Inc.: Cygwin
> 	`binmode'). There are various issues with Cygwin versions before
> 	1.3.2 - you are adviced to update to the latest release.
>
> -	Dynamic loading works as of cygwin-1.1.3 and binutils-20000722-1.
> -	It was not tested for earlier versions. This does not imply
> -	that every module will work. New completion and in
> -	particular zsh/zftp and zsh/mathfunc are known to work.
> +	Process substitution using <(...), >(...), =(...) may be
> +	problematic.  Different versions of zsh and Cygwin have
> +	a different mix of issues.
> +
> +	Problems handling subprocesses have been reported with Cygwin
> +	1.7.5.  It is not currently known how the problems split between
> +	Cygwin and zsh.
>
> 	Some of the tests in the Test subdirectory are known to fail:
> 	this is because the UNIX environment is not completely implemented.

--
Peter A. Castro <doctor@xxxxxxxxxxxx> or <Peter.Castro@xxxxxxxxxx>
	"Cats are just autistic Dogs" -- Dr. Tony Attwood