Subject: Re: pkg/12705: pkgsrc/www/mozilla for 0.8.1 fails to build on i388 1.5T
To: Frederick Bruckman <fb@enteract.com>
From: Greg A. Woods <woods@weird.com>
List: netbsd-bugs
Date: 05/05/2001 11:31:26
[ On Saturday, May 5, 2001 at 08:20:11 (-0500), Frederick Bruckman wrote: ]
> Subject: Re: pkg/12705: pkgsrc/www/mozilla for 0.8.1 fails to build on i388  1.5T
>
> Yes. I remember now. I had built mozilla from a cvs checkout in the
> past with --without-pthreads, because there's no advantage to using
> pthreads on a single processor system -- for the user, maybe it's
> easier for the developer. Therefore, if we're given the option to use
> no-pthreads alternative code, as a rule, we should take it.

If the code really does make good use of threads then it may perform
better on even a single processor system -- something that might be
important for some types of user interfaces.
 
In any case the build I did with pth didn't work anyway, at least not
with my 1.5T kernel.  The install didn't even finish because it hung
somewhere in "regchrome":

	[[....]]/scripts/moz-install
	*** Registering -chat handler.
	*** Registering x-application-irc handler.
	*** Registering irc protocol handler.
	registerSelf for remoteControl
	Runtime mismatch, so leaking context!
	Runtime mismatch, so leaking context!

 UID   PID  PPID CPU PRI NI  VSZ  RSS WCHAN  STAT TT    TIME COMMAND
   0 13507 13208  13   2  0 1052 6024 select S+   p3 0:01.09 ./regchrome 

The installed mozilla_bin fails to run too, hanging the same way
(i.e. silently and chewing very little CPU).

A ktrace of either shows that it gets quite some way along and then just
goes into a loop until you interrupt it:

  9277 mozilla-bin CALL  gettimeofday(0xbfbfcce8,0)
  9277 mozilla-bin RET   gettimeofday 0
  9277 mozilla-bin CALL  ioctl(0xa,FIONREAD,0xbfbfc4b8)
  9277 mozilla-bin RET   ioctl 0
  9277 mozilla-bin CALL  poll(0x822a360,0x3,0x9)
  9277 mozilla-bin RET   poll 0
  9277 mozilla-bin CALL  gettimeofday(0xbfbfcce8,0)
  9277 mozilla-bin RET   gettimeofday 0
  9277 mozilla-bin CALL  gettimeofday(0xbfbfcc28,0)
  9277 mozilla-bin RET   gettimeofday 0
  9277 mozilla-bin CALL  ioctl(0xa,FIONREAD,0xbfbfc3f8)
  9277 mozilla-bin RET   ioctl 0
  9277 mozilla-bin CALL  poll(0x822a360,0x3,0)
  9277 mozilla-bin RET   poll 0
  9277 mozilla-bin CALL  gettimeofday(0xbfbfcce8,0)
  9277 mozilla-bin RET   gettimeofday 0
  9277 mozilla-bin CALL  ioctl(0xa,FIONREAD,0xbfbfc4b8)
  9277 mozilla-bin RET   ioctl 0
  9277 mozilla-bin CALL  poll(0x822a360,0x3,0x9)
  9277 mozilla-bin PSIG  SIGINT SIG_DFL


> In any case, I suggest we add "CONFIGURE_ENV+= --without-pthreads" to
> the mix. This should build the same way for folks who have managed to
> build it already. Agreed?

I'm re-starting my build now with "CONFIGURE_ARGS+= --without-pthreads"
[[ s/_ENV/_ARGS/ :-) ]]
 
I see that the configure scripts still tested for (and found)
pthread_create(), but it seems it successfully built that library it was
tripping over before without -lpthread this time so we'll see how this
one runs in a few hours....

> [I did leave a build running at home, but I'm afraid the partition on
> which ${WRKOBJDIR} is only ~500mb, so I half expect to find a "disk
> full" when I return.]

If there was that much space left after the extract was done then you
might be OK....

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>     <woods@robohack.ca>
Planix, Inc. <woods@planix.com>;   Secrets of the Weird <woods@weird.com>