revno 101399 fails to build

revno 101399 fails to build

[joakim]

...
In toplevel form:
calc/calc-help.el:30:1:Error: Required feature `calc-ext' was not provided
make[3]: *** [calc/calc-help.elc] Error 1
make[3]: *** Waiting for unfinished jobs....
Wrote /var/lib/hudson/jobs/emacs-trunk-daily/workspace/lisp/calc/calc-graph.elc
Wrote /var/lib/hudson/jobs/emacs-trunk-daily/workspace/lisp/calc/calc-incom.elc
make[3]: Leaving directory `/var/lib/hudson/jobs/emacs-trunk-daily/workspace/lisp'
make[2]: *** [compile-main] Error 2
make[2]: Leaving directory `/var/lib/hudson/jobs/emacs-trunk-daily/workspace/lisp'
make[1]: *** [lisp] Error 2
make[1]: Leaving directory `/var/lib/hudson/jobs/emacs-trunk-daily/workspace'
make: *** [bootstrap] Error 2

-- 
Joakim Verona

Re: revno 101399 fails to build

[Glenn Morris]

joakim@verona.se wrote:

> calc/calc-help.el:30:1:Error: Required feature `calc-ext' was not provided
> make[3]: *** [calc/calc-help.elc] Error 1
> make[3]: *** Waiting for unfinished jobs....

Builds fine for me. I imagine this was a transient parallel build
failure, as can happen from time to time. Meditate on success and try
again.

http://debbugs.gnu.org/cgi/bugreport.cgi?bug=6858

Re: revno 101399 fails to build

[joakim]

Glenn Morris <rgm@gnu.org> writes:

> joakim@verona.se wrote:
>
>> calc/calc-help.el:30:1:Error: Required feature `calc-ext' was not provided
>> make[3]: *** [calc/calc-help.elc] Error 1
>> make[3]: *** Waiting for unfinished jobs....
>
> Builds fine for me. I imagine this was a transient parallel build
> failure, as can happen from time to time. Meditate on success and try
> again.
>
> http://debbugs.gnu.org/cgi/bugreport.cgi?bug=6858

Ok thanks for this information.

I now removed the "-j4" flag I had in my emacs build server.

Funnily though, I had 2 jobs building emacs, both failed the same way.
The chance of that ought to be rather low if its a race condition.

-- 
Joakim Verona

Re: revno 101399 fails to build

[Andreas Schwab]

joakim@verona.se writes:

> ...
> In toplevel form:
> calc/calc-help.el:30:1:Error: Required feature `calc-ext' was not provided

Do you build with more that one job in parallel?  In this case it can
happen that one emacs process writes out the elc file while another one
is reading the same (partial) file at the same time.  To fix that the
creation of the elc file need to be made atomic.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: revno 101399 fails to build

[joakim]

Andreas Schwab <schwab@linux-m68k.org> writes:

> joakim@verona.se writes:
>
>> ...
>> In toplevel form:
>> calc/calc-help.el:30:1:Error: Required feature `calc-ext' was not provided
>
> Do you build with more that one job in parallel?  In this case it can
> happen that one emacs process writes out the elc file while another one
> is reading the same (partial) file at the same time.  To fix that the
> creation of the elc file need to be made atomic.

Ok, given the current infrastructure, what combination of build commands
would be as reliable as possible and reasonably fast?

is it possible to do, in pseudo-code:

make -j4 all-c-targets && \
make elc-files && \
make -j4 all-other-stuff-requiring-c-targets-and-elc-files

This is for a build server, and I want very reliable build results.

I have/want nightly, slow,  code quality builds, and continus, fast, builds
polling bzr.


> Andreas.

-- 
Joakim Verona

Re: revno 101399 fails to build

[Stephen J. Turnbull]

joakim@verona.se writes:

 > Funnily though, I had 2 jobs building emacs, both failed the same
 > way.  The chance of that ought to be rather low if its a race
 > condition.

(1) Not if they're both using the same files!  Offhand, I don't know
    about Emacs's build process, but in XEmacs the elc building
    process is very lazy since there is normally no dependency on
    platform or build in the elcs, only on changes in the els -- we
    symlink the whole lisp directory into the build directory, and
    build the elcs in-place.

(2) Since race conditions depend on timing (otherwise they'd be easy
    to diagnose and fix plain ol' bugs!), there may be something about
    the build host that frequently creates the right timing.  Or
    possibly building from the same source directory does that on your
    host.

Re: revno 101399 fails to build

[Eli Zaretskii]

> From: Andreas Schwab <schwab@linux-m68k.org>
> Date: Fri, 10 Sep 2010 10:37:43 +0200
> Cc: Emacs developers <emacs-devel@gnu.org>
> 
> joakim@verona.se writes:
> 
> > ...
> > In toplevel form:
> > calc/calc-help.el:30:1:Error: Required feature `calc-ext' was not provided
> 
> Do you build with more that one job in parallel?  In this case it can
> happen that one emacs process writes out the elc file while another one
> is reading the same (partial) file at the same time.  To fix that the
> creation of the elc file need to be made atomic.

Or specify explicit dependencies in the Makefile.  Or get Emacs to
generate such dependencies given the `require's.

Re: revno 101399 fails to build

[Eli Zaretskii]

> From: "Stephen J. Turnbull" <stephen@xemacs.org>
> Date: Sat, 11 Sep 2010 13:58:46 +0900
> Cc: Emacs developers <emacs-devel@gnu.org>
> 
>                                            Offhand, I don't know
>     about Emacs's build process, but in XEmacs the elc building
>     process is very lazy since there is normally no dependency on
>     platform or build in the elcs, only on changes in the els -- we
>     symlink the whole lisp directory into the build directory

Which doesn't work on Windows.

Re: revno 101399 fails to build

[Stefan Monnier]

>> > ...
>> > In toplevel form:
>> > calc/calc-help.el:30:1:Error: Required feature `calc-ext' was not provided
>> 
>> Do you build with more that one job in parallel?  In this case it can
>> happen that one emacs process writes out the elc file while another one
>> is reading the same (partial) file at the same time.  To fix that the
>> creation of the elc file need to be made atomic.

> Or specify explicit dependencies in the Makefile.  Or get Emacs to
> generate such dependencies given the `require's.

BTW, I did play with that a long time ago and the result was that it
works to some extent but it's not enough: many dependencies come from
autoloaded functions/macros, there are circular require-dependencies,
many dependencies come from preloaded files without any `require'
in sight.
Maybe a more promising direction is to let the byte-compiler emit
dependencies as it goes, just like gcc can do.


        Stefan

Re: revno 101399 fails to build

[Glenn Morris]

>>  To fix that the creation of the elc file need to be made atomic.

Hopefully done now, BTW.

Re: revno 101399 fails to build

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Cc: Andreas Schwab <schwab@linux-m68k.org>,  joakim@verona.se,  emacs-devel@gnu.org
> Date: Mon, 13 Sep 2010 17:52:59 +0200
> 
> Maybe a more promising direction is to let the byte-compiler emit
> dependencies as it goes, just like gcc can do.

That's what I had in mind, yes.

Re: revno 101399 fails to build

[Stephen J. Turnbull]

Eli Zaretskii writes:
 > > From: "Stephen J. Turnbull" <stephen@xemacs.org>
 > > Date: Sat, 11 Sep 2010 13:58:46 +0900
 > > Cc: Emacs developers <emacs-devel@gnu.org>
 > > 
 > >                                            Offhand, I don't know
 > >     about Emacs's build process, but in XEmacs the elc building
 > >     process is very lazy since there is normally no dependency on
 > >     platform or build in the elcs, only on changes in the els --
 > >     we symlink the whole lisp directory into the build directory
 > 
 > Which doesn't work on Windows.

My understanding was that we were discussing a build on a POSIX-ish
platform.  If not, my bad.