Bootstrap failure using 'make -j4' [Cygwin]

Bootstrap failure using 'make -j4' [Cygwin]

[Angelo Graziosi]

Today, trying to bootstrap Emacs trunk (rev. 100406), using 'make -j4', 
fails always in the same manner:

[...]
gcc -c  -Demacs -DHAVE_CONFIG_H  -I. -I/tmp/emacs/src 
-I/usr/include/gtk-2.0 -I/usr/lib/gtk-2.0/include -I/usr/include/atk-1.0 
-I/usr/include/cairo -I/usr/include/pango-1.0 -I/usr/include/glib-2.0 
-I/usr/lib/glib-2.0/include -I/usr/include/pixman-1 
-I/usr/include/freetype2 -I/usr/include/libpng12 
-I/usr/include/freetype2     -I/usr/include/librsvg-2 
-I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include 
-I/usr/include/gtk-2.0 -I/usr/include/cairo -I/usr/include/pixman-1 
-I/usr/include/freetype2 -I/usr/include/libpng12 
-I/usr/include/dbus-1.0 -I/usr/lib/dbus-1.0/include   -DORBIT2=1 
-D_REENTRANT -I/usr/include/gconf/2 -I/usr/include/orbit-2.0 
-I/usr/include/dbus-1.0 -I/usr/lib/dbus-1.0/include 
-I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include   -g -O2 
-Wdeclaration-after-statement -Wno-pointer-sign       -MMD -MF 
deps/frame.d /tmp/emacs/src/frame.c
make[2]: *** [dispnew.o] Error 1
make[2]: *** Waiting for unfinished jobs....
make[2]: Leaving directory `/tmp/emacs/build/src'
make[3]: Leaving directory `/tmp/emacs/build/doc/lispref'
make[1]: *** [src] Error 2
make[1]: *** Waiting for unfinished jobs....
(cd doc/lispintro; /usr/bin/make -w --jobserver-fds=3,4 - 
--jobserver-fds=3,4 - --jobserver-fds=3,4 - --jobserver-fds=3,4 -j info)
make[3]: Entering directory `/tmp/emacs/build/doc/lispintro'
cd /tmp/emacs/doc/lispintro; makeinfo emacs-lisp-intro.texi -o 
../../info/eintr
make[3]: Leaving directory `/tmp/emacs/build/doc/lispintro'
make[2]: Leaving directory `/tmp/emacs/build'
make[1]: Leaving directory `/tmp/emacs/build'
make: *** [bootstrap] Error 2

This happens for the GTK build on Cygwin.

I have done many tries... It looks as if, with 'make -j4', it avoids to 
compile dispnew.c. Indeed I obtain

$ grep dispnew build-emacs-gtk-24.0.50-r100405.log
make[2]: *** [dispnew.o] Error 1

Today I can bootstrap only using 'make -j2' :(

Ciao,
Angelo.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> Date: Fri, 21 May 2010 16:53:15 +0200
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> 
> Today, trying to bootstrap Emacs trunk (rev. 100406), using 'make -j4', 
> fails always in the same manner:

Thanks for the report, but where's the relevant data?  It fails
compiling dispnew.c, but the portion of the build you show relates to
frame.c, not to dispnew.c.  Please show more.

> I have done many tries... It looks as if, with 'make -j4', it avoids to 
> compile dispnew.c. Indeed I obtain
> 
> $ grep dispnew build-emacs-gtk-24.0.50-r100405.log
> make[2]: *** [dispnew.o] Error 1

It doesn't avoid.  It tries to compile it, but the compilation fails
for some reason.  The question is what is the reason.  Seeing the
compiler command and its output might help.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Angelo Graziosi]

Il 21/05/2010 18.16, Eli Zaretskii ha scritto:
>> Date: Fri, 21 May 2010 16:53:15 +0200
>> From: Angelo Graziosi<angelo.graziosi@alice.it>
>>
>> Today, trying to bootstrap Emacs trunk (rev. 100406), using 'make -j4',
>> fails always in the same manner:
>
> Thanks for the report, but where's the relevant data?  It fails
> compiling dispnew.c, but the portion of the build you show relates to
> frame.c, not to dispnew.c.  Please show more.
>
>> I have done many tries... It looks as if, with 'make -j4', it avoids to
>> compile dispnew.c. Indeed I obtain
>>
>> $ grep dispnew build-emacs-gtk-24.0.50-r100405.log
>> make[2]: *** [dispnew.o] Error 1
>
> It doesn't avoid.  It tries to compile it, but the compilation fails
> for some reason.  The question is what is the reason.  Seeing the
> compiler command and its output might help.

Hmm... perhaps I wasn't clear.

In the build log there is *not* track of the command to compile 
dispnew.c: as you see 'grep' finds only

make[2]: *** [dispnew.o] Error 1

But this happens *only* with 'make -j4'! I one uses 'make -j2' the 
bootstrap is completed and works, and 'dispnew.c' is compiled 
successfully with:

gcc -c  -Demacs -DHAVE_CONFIG_H  -I. -I/tmp/emacs/src 
-I/usr/include/gtk-2.0 -I/usr/lib/gtk-2.0/include -I/usr/include/atk-1.0 
-I/usr/include/cairo -I/usr/include/pango-1.0 -I/usr/include/glib-2.0 
-I/usr/lib/glib-2.0/include -I/usr/include/pixman-1 
-I/usr/include/freetype2 -I/usr/include/libpng12 
-I/usr/include/freetype2     -I/usr/include/librsvg-2 
-I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include 
-I/usr/include/gtk-2.0 -I/usr/include/cairo -I/usr/include/pixman-1 
-I/usr/include/freetype2 -I/usr/include/libpng12 
-I/usr/include/dbus-1.0 -I/usr/lib/dbus-1.0/include   -DORBIT2=1 
-D_REENTRANT -I/usr/include/gconf/2 -I/usr/include/orbit-2.0 
-I/usr/include/dbus-1.0 -I/usr/lib/dbus-1.0/include 
-I/usr/include/glib-2.0 -I/usr/lib/glib-2.0/include   -g -O2 
-Wdeclaration-after-statement -Wno-pointer-sign       -MMD -MF 
deps/dispnew.d /tmp/emacs/src/dispnew.c


I know that with '-j4' some time the build can fail, mainly if the CPU 
is doing other things, and re-trying often solves. But in this case the 
CPU was working only for Emacs bootstrap and trying many times did not 
solve: it fails always in the same manner. Only using '-j2' did solve!

It is at least two years I use 'make -j4' on Cygwin to bootstrap Emacs 
without problems (indeed it worked a week ago...)

Ciao,
Angelo.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> Date: Fri, 21 May 2010 18:57:41 +0200
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Cc: emacs-devel@gnu.org
> 
> > It doesn't avoid.  It tries to compile it, but the compilation fails
> > for some reason.  The question is what is the reason.  Seeing the
> > compiler command and its output might help.
> 
> Hmm... perhaps I wasn't clear.
> 
> In the build log there is *not* track of the command to compile 
> dispnew.c: as you see 'grep' finds only
> 
> make[2]: *** [dispnew.o] Error 1

What happens if you say "make -j4 -d 2>&1 | tee foo"?  Do you see any
references to dispnew in the file `foo' created by this command?

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> Date: Fri, 21 May 2010 18:57:41 +0200
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> Cc: emacs-devel@gnu.org
> 
> In the build log there is *not* track of the command to compile 
> dispnew.c: as you see 'grep' finds only
> 
> make[2]: *** [dispnew.o] Error 1

I cannot reproduce this on GNU/Linux with "make -j4 bootstrap".  How
many CPU cores do you have on that machine?

If nothing else gives a clue, post the entire transcript of the
"configure" and "make -j4 bootstrap" steps.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> Date: Fri, 21 May 2010 20:59:44 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: emacs-devel@gnu.org
> 
> > Date: Fri, 21 May 2010 18:57:41 +0200
> > From: Angelo Graziosi <angelo.graziosi@alice.it>
> > Cc: emacs-devel@gnu.org
> > 
> > In the build log there is *not* track of the command to compile 
> > dispnew.c: as you see 'grep' finds only
> > 
> > make[2]: *** [dispnew.o] Error 1
> 
> I cannot reproduce this on GNU/Linux with "make -j4 bootstrap".

In fact, I cannot reproduce this with -jN for N in [2..6].  With -j6,
I get a failure messages like this

    make[3]: vfork: Resource temporarily unavailable
    /bin/sh: Cannot fork

which I think mean I'm trying to use more resources than I'm allowed
to on this machine.  But that's way after bootstrap-emacs was linked,
including compiling dispnew.c.

Do you see any other unusual messages in the transcript?

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Angelo Graziosi]

[-- Attachment #1: Type: text/plain, Size: 989 bytes --]

Il 21/05/2010 19.59, Eli Zaretskii ha scritto:
>> Date: Fri, 21 May 2010 18:57:41 +0200
>> From: Angelo Graziosi<>
>> Cc:
>>
>> In the build log there is *not* track of the command to compile
>> dispnew.c: as you see 'grep' finds only
>>
>> make[2]: *** [dispnew.o] Error 1
>
> I cannot reproduce this on GNU/Linux with "make -j4 bootstrap".

Indeed. It bootstraps fine on GNU/Linux Kubuntu 10.04

> How many CPU cores do you have on that machine?

It is an AMD Athlon 64 X2 Dual Core Processor 3800+ 2.03 GHz, 1.75GB RAM

> If nothing else gives a clue, post the entire transcript of the
> "configure" and "make -j4 bootstrap" steps.

For completeness I have attached the full build log (40K, 8K compressed).

With 'make -j4 -d 2>&1 | tee foo', as you suggested, I cannot reproduce 
the problem: it seems to bootstrap completely (at least emacs.exe is 
built and dispnew.c compiled).

My build script use

${source_dir}/configure --prefix=${prefix_dir}
make -j4 bootstrap


Ciao,
Angelo.

[-- Attachment #2: build-emacs-gtk-24.0.50-r100410.log.bz2 --]
[-- Type: application/octet-stream, Size: 7901 bytes --]

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> Date: Sat, 22 May 2010 00:45:31 +0200
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> CC: emacs-devel@gnu.org
> 
> > How many CPU cores do you have on that machine?
> 
> It is an AMD Athlon 64 X2 Dual Core Processor 3800+ 2.03 GHz, 1.75GB RAM

If you have only 2 cores, then -j4 is not really useful, is it?  You
will never have more than 2 jobs running at the same time.  I think
the usual rule of thumb is to use 1.5 times the number of cores as the
argument to -j.  In your case, that would be -j3.

> > If nothing else gives a clue, post the entire transcript of the
> > "configure" and "make -j4 bootstrap" steps.
> 
> For completeness I have attached the full build log (40K, 8K compressed).

Thanks.  I think your problem begins with this error message I see in
the log:

   mkdir: impossibile creare la directory "deps": File exists

This comes from the following fragment in src/Makefile.in:

    .c.o:
	    @$(MKDEPDIR)
	    $(CC) -c $(CPPFLAGS) $(ALL_CFLAGS) $<

where MKDEPDIR expands to this:

    test -d ${DEPDIR} || mkdir ${DEPDIR}

Now, if more than one .c.o rule tries to create the `deps'
subdirectory at the same time, one of them might lose due to a race
condition, and fail the entire compilation.

Could you please try changing this line:

	    @$(MKDEPDIR)
into
	    -@$(MKDEPDIR)

and see if that makes the problem go away?

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Glenn Morris]

Eli Zaretskii wrote:

> where MKDEPDIR expands to this:
>
>     test -d ${DEPDIR} || mkdir ${DEPDIR}
>
> Now, if more than one .c.o rule tries to create the `deps'
> subdirectory at the same time, one of them might lose due to a race
> condition, and fail the entire compilation.

Please note I changed this earlier today for this reason; but in a
different way to your suggestion.

> Could you please try changing this line:
>
> 	    @$(MKDEPDIR)
> into
> 	    -@$(MKDEPDIR)

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Glenn Morris]

PS it's my fault for overlooking the "-" in the previous "@-" version.
Hard to spot!

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> Cc: Angelo Graziosi <angelo.graziosi@alice.it>,  emacs-devel@gnu.org
> From: Glenn Morris <rgm@gnu.org>
> Date: Sat, 22 May 2010 03:16:00 -0400
> 
> Eli Zaretskii wrote:
> 
> > where MKDEPDIR expands to this:
> >
> >     test -d ${DEPDIR} || mkdir ${DEPDIR}
> >
> > Now, if more than one .c.o rule tries to create the `deps'
> > subdirectory at the same time, one of them might lose due to a race
> > condition, and fail the entire compilation.
> 
> Please note I changed this earlier today for this reason

Changed how?  I don't see any changes that would solve this.  With the
current trunk, running configure on GNU/Linux still produces
src/Makefile with these fragments:

    MKDEPDIR=test -d ${DEPDIR} || mkdir ${DEPDIR}
    ...
    .c.o:
    	@$(MKDEPDIR)
    	$(CC) -c $(CPPFLAGS) $(ALL_CFLAGS) $<


What am I missing?

> but in a different way to your suggestion.

That wasn't a suggestion for a change in src/Makefile.in, just a means
to make sure Angelo's problem is due to this race, and not to
something else.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Jan Djärv]

Eli Zaretskii skrev 2010-05-22 08.57:

> Thanks.  I think your problem begins with this error message I see in
> the log:
>
>     mkdir: impossibile creare la directory "deps": File exists
>
> This comes from the following fragment in src/Makefile.in:
>
>      .c.o:
> 	    @$(MKDEPDIR)
> 	    $(CC) -c $(CPPFLAGS) $(ALL_CFLAGS) $<
>
> where MKDEPDIR expands to this:
>
>      test -d ${DEPDIR} || mkdir ${DEPDIR}
>
> Now, if more than one .c.o rule tries to create the `deps'
> subdirectory at the same time, one of them might lose due to a race
> condition, and fail the entire compilation.
>
> Could you please try changing this line:
>
> 	    @$(MKDEPDIR)
> into
> 	    -@$(MKDEPDIR)
>
> and see if that makes the problem go away?

A better solution is probably to replace

       test -d ${DEPDIR} || mkdir ${DEPDIR}

with

       mkdir -p ${DEPDIR}

	Jan D.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Andreas Schwab]

Eli Zaretskii <eliz@gnu.org> writes:

> What am I missing?

You need to regenerate configure.

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: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> Date: Sat, 22 May 2010 13:00:53 +0200
> From: Jan Djärv <jan.h.d@swipnet.se>
> CC: Angelo Graziosi <angelo.graziosi@alice.it>, emacs-devel@gnu.org
> 
> A better solution is probably to replace
> 
>        test -d ${DEPDIR} || mkdir ${DEPDIR}
> 
> with
> 
>        mkdir -p ${DEPDIR}

I think the -p switch to mkdir is not portable.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> From: Andreas Schwab <schwab@linux-m68k.org>
> Cc: Glenn Morris <rgm@gnu.org>,  emacs-devel@gnu.org,  angelo.graziosi@alice.it
> Date: Sat, 22 May 2010 13:20:50 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > What am I missing?
> 
> You need to regenerate configure.

Right, I assumed Glenn did.

Thanks.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Jan Djärv]

Eli Zaretskii skrev 2010-05-22 14.46:
>> Date: Sat, 22 May 2010 13:00:53 +0200
>> From: Jan Djärv<jan.h.d@swipnet.se>
>> CC: Angelo Graziosi<angelo.graziosi@alice.it>, emacs-devel@gnu.org
>>
>> A better solution is probably to replace
>>
>>         test -d ${DEPDIR} || mkdir ${DEPDIR}
>>
>> with
>>
>>         mkdir -p ${DEPDIR}
>
> I think the -p switch to mkdir is not portable.
>

I has been in The Single UNIX ® Specification since 1997 at least.  I think it 
was in POSIX 2 as well (1990), but I lost it, so I can't check.

Do you know any system that doesn't have it?

	Jan D.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[David Engster]

Jan Djärv writes:
> Eli Zaretskii skrev 2010-05-22 14.46:
>>> Date: Sat, 22 May 2010 13:00:53 +0200
>>> From: Jan Djärv<jan.h.d@swipnet.se>
>>> CC: Angelo Graziosi<angelo.graziosi@alice.it>, emacs-devel@gnu.org
>>>
>>> A better solution is probably to replace
>>>
>>>         test -d ${DEPDIR} || mkdir ${DEPDIR}
>>>
>>> with
>>>
>>>         mkdir -p ${DEPDIR}
>>
>> I think the -p switch to mkdir is not portable.
>>
>
> I has been in The Single UNIX ® Specification since 1997 at least.  I
> think it was in POSIX 2 as well (1990), but I lost it, so I can't
> check.

That may be, but 'mkdir -p' is not guaranteed to be thread-safe.

-David

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> Date: Sat, 22 May 2010 15:00:01 +0200
> From: Jan Djärv <jan.h.d@swipnet.se>
> CC: emacs-devel@gnu.org, angelo.graziosi@alice.it
> 
> > I think the -p switch to mkdir is not portable.
> >
> 
> I has been in The Single UNIX ® Specification since 1997 at least.  I think it 
> was in POSIX 2 as well (1990), but I lost it, so I can't check.
> 
> Do you know any system that doesn't have it?

Not by name, no.  But this is in the GNU Coding Statndards:

       Generally, stick to the widely-supported (usually POSIX-specified)
    options and features of these programs.  For example, don't use `mkdir
    -p', convenient as it may be, because a few systems don't support it at
    all and with others, it is not safe for parallel execution.

If this is no longer a consideration, perhaps we should ask Richard to
amend standards.texi.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Angelo Graziosi]

The problem seems fixed in rev 100412, 'make -j4' works just fine 
(thanks for having clarified the cores usage of CPU)

And...

Il 22/05/2010 8.57, Eli Zaretskii ha scritto:
> Thanks.  I think your problem begins with this error message I see in
> the log:
>
>     mkdir: impossibile creare la directory "deps": File exists

for completeness, I noticed this error on GNU/Linux because it caused 
the build to be stopped... But there I thought this was caused by some 
garbage from a previous build and, really, after deleting the build 
tree, I was able to bootstrap...

Ciao,
Angelo.

PS. Should I test what follow in any case?

> This comes from the following fragment in src/Makefile.in:
>
>      .c.o:
> 	    @$(MKDEPDIR)
> 	    $(CC) -c $(CPPFLAGS) $(ALL_CFLAGS) $<
>
> where MKDEPDIR expands to this:
>
>      test -d ${DEPDIR} || mkdir ${DEPDIR}
>
> Now, if more than one .c.o rule tries to create the `deps'
> subdirectory at the same time, one of them might lose due to a race
> condition, and fail the entire compilation.
>
> Could you please try changing this line:
>
> 	    @$(MKDEPDIR)
> into
> 	    -@$(MKDEPDIR)
>
> and see if that makes the problem go away?

Re: Bootstrap failure using 'make -j4' [Cygwin]

[David Engster]

Eli Zaretskii writes:
>> Date: Sat, 22 May 2010 15:00:01 +0200
>> From: Jan Djärv <jan.h.d@swipnet.se>
>> CC: emacs-devel@gnu.org, angelo.graziosi@alice.it
>> 
>> > I think the -p switch to mkdir is not portable.
>> >
>> 
>> I has been in The Single UNIX ® Specification since 1997 at least.  I think it 
>> was in POSIX 2 as well (1990), but I lost it, so I can't check.
>> 
>> Do you know any system that doesn't have it?
>
> Not by name, no.  But this is in the GNU Coding Statndards:
>
>        Generally, stick to the widely-supported (usually POSIX-specified)
>     options and features of these programs.  For example, don't use `mkdir
>     -p', convenient as it may be, because a few systems don't support it at
>     all and with others, it is not safe for parallel execution.
>
> If this is no longer a consideration, perhaps we should ask Richard to
> amend standards.texi.

In autoconf, Solaris 8 is mentioned with a thread-unsafe 'mkdir -p', so
this is still an issue. Autoconf's MKDIR_P falls back to 'install-sh -d'
when GNU mkdir isn't available.

-David

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Jan Djärv]

David Engster skrev 2010-05-22 15.06:
> Jan Djärv writes:
>> Eli Zaretskii skrev 2010-05-22 14.46:
>>>> Date: Sat, 22 May 2010 13:00:53 +0200
>>>> From: Jan Djärv<jan.h.d@swipnet.se>
>>>> CC: Angelo Graziosi<angelo.graziosi@alice.it>, emacs-devel@gnu.org
>>>>
>>>> A better solution is probably to replace
>>>>
>>>>          test -d ${DEPDIR} || mkdir ${DEPDIR}
>>>>
>>>> with
>>>>
>>>>          mkdir -p ${DEPDIR}
>>>
>>> I think the -p switch to mkdir is not portable.
>>>
>>
>> I has been in The Single UNIX ® Specification since 1997 at least.  I
>> think it was in POSIX 2 as well (1990), but I lost it, so I can't
>> check.
>
> That may be, but 'mkdir -p' is not guaranteed to be thread-safe.

That has nothing to do with this situation.  The point of -p here is that 
mkdir shall not output an error if the directory exist.  When creating 
multiple directories, then -p is not thread safe (i.e. two mkdir:s may be 
creating directories in the tree), but this is not the case here.

	Jan D.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Jan Djärv]

Eli Zaretskii skrev 2010-05-22 15.21:
>> Date: Sat, 22 May 2010 15:00:01 +0200
>> From: Jan Djärv<jan.h.d@swipnet.se>
>> CC: emacs-devel@gnu.org, angelo.graziosi@alice.it
>>
>>> I think the -p switch to mkdir is not portable.
>>>
>>
>> I has been in The Single UNIX ® Specification since 1997 at least.  I think it
>> was in POSIX 2 as well (1990), but I lost it, so I can't check.
>>
>> Do you know any system that doesn't have it?
>
> Not by name, no.  But this is in the GNU Coding Statndards:
>
>         Generally, stick to the widely-supported (usually POSIX-specified)
>      options and features of these programs.  For example, don't use `mkdir
>      -p', convenient as it may be, because a few systems don't support it at
>      all and with others, it is not safe for parallel execution.
>
> If this is no longer a consideration, perhaps we should ask Richard to
> amend standards.texi.
>

Well, -p is in POSIX now, so I guess the standards needs to be updated on that 
point.  But it is still not parallel safe.

Emacs uses prototypes in C without macros now, and that is something the GNU 
coding standards says not to do.  So I really don't know how relevant the GNU 
Coding standards are for Emacs developemnt.

	Jan D.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Jan Djärv]

David Engster skrev 2010-05-22 15.41:

>
> In autoconf, Solaris 8 is mentioned with a thread-unsafe 'mkdir -p', so
> this is still an issue. Autoconf's MKDIR_P falls back to 'install-sh -d'
> when GNU mkdir isn't available.
>

Thread safety isn't an issue for the problem at hand.  Anyway, install-sh -d 
is not parallel safe either.

	Jan D.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> Date: Sat, 22 May 2010 15:35:34 +0200
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> CC: emacs-devel@gnu.org
> 
> PS. Should I test what follow in any case?

No need now.  Thanks.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> Date: Sat, 22 May 2010 15:54:29 +0200
> From: Jan Djärv <jan.h.d@swipnet.se>
> CC: angelo.graziosi@alice.it, emacs-devel@gnu.org
> 
> >         Generally, stick to the widely-supported (usually POSIX-specified)
> >      options and features of these programs.  For example, don't use `mkdir
> >      -p', convenient as it may be, because a few systems don't support it at
> >      all and with others, it is not safe for parallel execution.
> >
> > If this is no longer a consideration, perhaps we should ask Richard to
> > amend standards.texi.
> >
> 
> Well, -p is in POSIX now, so I guess the standards needs to be updated on that 
> point.  But it is still not parallel safe.

If it's not parallel safe, it might be a problem in this case, since
multiple .c.o rules could invoke it in parallel under "make -j".

> Emacs uses prototypes in C without macros now, and that is something the GNU 
> coding standards says not to do.  So I really don't know how relevant the GNU 
> Coding standards are for Emacs developemnt.

They are relevant, and the document is actively maintained.  Any
anachronisms you find there I suggest to report to Richard.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Angelo Graziosi]

Il 22/05/2010 8.57, Eli Zaretskii ha scritto:
> If you have only 2 cores, then -j4 is not really useful, is it?  You
> will never have more than 2 jobs running at the same time.  I think
> the usual rule of thumb is to use 1.5 times the number of cores as the
> argument to -j.  In your case, that would be -j3.

Just for completeness, whenever I use -j2, -j3, -j4, I find always this 
warning:

[...]
if [ -x ./config.status ]; then           \
	    ./config.status;                      \
	else                                      \
	    ./configure --enable-maintainer-mode; \
	fi
config.status: creating Makefile
config.status: creating lib-src/Makefile
config.status: creating oldXMenu/Makefile
config.status: creating doc/emacs/Makefile
config.status: creating doc/misc/Makefile
config.status: creating doc/lispintro/Makefile
config.status: creating doc/lispref/Makefile
config.status: creating src/Makefile.c
config.status: creating lwlib/Makefile
config.status: creating lisp/Makefile
config.status: creating leim/Makefile
config.status: creating src/config.h
config.status: executing default commands
creating src/epaths.h
make[1]: Entering directory `/tmp/emacs/build'
make[1]: warning: jobserver unavailable: using -j1.  Add `+' to parent 
make rule.
make[1]: Leaving directory `/tmp/emacs/build'
[...]
=======================

Ciao,
Angelo.

Re: Bootstrap failure using 'make -j4' [Cygwin]

[Eli Zaretskii]

> Date: Sun, 23 May 2010 13:51:22 +0200
> From: Angelo Graziosi <angelo.graziosi@alice.it>
> CC: emacs-devel@gnu.org
> 
> make[1]: Entering directory `/tmp/emacs/build'
> make[1]: warning: jobserver unavailable: using -j1.  Add `+' to parent 
> make rule.
> make[1]: Leaving directory `/tmp/emacs/build'

I think this is normal: config.status runs "make epaths-force", but
since this is not a sub-make of the main Make, the -j switch is
disabled.