Symbol's chain of function indirections contains a loop

Symbol's chain of function indirections contains a loop

[Eli Zaretskii]

Bootstrapping the current trunk shows a lot of error messages like
this:

  Compiling emacs-lisp/copyright.el

  In toplevel form:
  emacs-lisp/copyright.el:346:1:Error: Symbol's chain of function indirections contains a loop: vc-update

Some of the messages are about vc-pull.

The bootstrap hangs completely when it comes to compile vc.el.

Anyone else sees this?

Re: Symbol's chain of function indirections contains a loop

[Christoph]

On 1/29/2011 9:19 AM, Eli Zaretskii wrote:

> Some of the messages are about vc-pull.
> The bootstrap hangs completely when it comes to compile vc.el.
> Anyone else sees this?

Yes, make bootstrap hangs for me on the following line:

Wrote d:/devel/emacs/emacs-bzr/trunk/lisp/vc/vc-svn.elc

Christoph

Re: Symbol's chain of function indirections contains a loop

[Tassilo Horn]

Christoph <cschol2112@googlemail.com> writes:

>> Some of the messages are about vc-pull.
>> The bootstrap hangs completely when it comes to compile vc.el.
>> Anyone else sees this?
>
> Yes, make bootstrap hangs for me on the following line:
>
> Wrote d:/devel/emacs/emacs-bzr/trunk/lisp/vc/vc-svn.elc

Mine went through without a problem an a GNU/Linux bug.  Maybe it's some
windows-only issue?

Bye,
Tassilo

Re: Symbol's chain of function indirections contains a loop

[Christoph]

On 01/29/2011 10:44 AM, Tassilo Horn wrote:

> Mine went through without a problem an a GNU/Linux bug.  Maybe it's some
> windows-only issue?

Looks like it. I just tried it on my GNU/Linux box (Ubuntu 10.10, 
compiling with recent trunk of clang) and it indeed succeeded without 
problem.

Christoph

Re: Symbol's chain of function indirections contains a loop

[Eli Zaretskii]

> From: Tassilo Horn <tassilo@member.fsf.org>
> Cc: Eli Zaretskii <eliz@gnu.org>,  emacs-devel@gnu.org
> Date: Sat, 29 Jan 2011 18:44:35 +0100
> 
> Christoph <cschol2112@googlemail.com> writes:
> 
> >> Some of the messages are about vc-pull.
> >> The bootstrap hangs completely when it comes to compile vc.el.
> >> Anyone else sees this?
> >
> > Yes, make bootstrap hangs for me on the following line:
> >
> > Wrote d:/devel/emacs/emacs-bzr/trunk/lisp/vc/vc-svn.elc
> 
> Mine went through without a problem an a GNU/Linux bug.  Maybe it's some
> windows-only issue?

Maybe, but I'd appreciate any ideas how this could happen.

For now, I've narrowed this down to the following command:

  "./../bin/emacs.exe" -batch --no-site-file --no-site-lisp -l loaddefs  -f batch-byte-compile-if-not-done expand.el

When this command is issued in the lisp/ subdirectory, it fails with
emacs.exe that is the one produced by the first build during the
bootstrap (see the `bootstrap-temacs' and `bootstrap-emacs' targets in
src/makefile.w32-in), but succeeds if I do the same with an Emacs
binary that is built normally.  The only difference between these two
binaries is that the former is built with -DPURESIZE=5000000, while
the latter uses PURESIZE defined on puresize.h.

Re: Symbol's chain of function indirections contains a loop

[Lennart Borgman]

On Sat, Jan 29, 2011 at 7:54 PM, Eli Zaretskii <eliz@gnu.org> wrote:
>> >
>> > Yes, make bootstrap hangs for me on the following line:
>> >
>> > Wrote d:/devel/emacs/emacs-bzr/trunk/lisp/vc/vc-svn.elc

It hangs there for me too.

Re: Symbol's chain of function indirections contains a loop

[Lennart Borgman]

On Sat, Jan 29, 2011 at 10:00 PM, Lennart Borgman
<lennart.borgman@gmail.com> wrote:
> On Sat, Jan 29, 2011 at 7:54 PM, Eli Zaretskii <eliz@gnu.org> wrote:
>>> >
>>> > Yes, make bootstrap hangs for me on the following line:
>>> >
>>> > Wrote d:/devel/emacs/emacs-bzr/trunk/lisp/vc/vc-svn.elc
>
> It hangs there for me too.

I have no idea of what is going on, but I notice these defaliases

./ldefs-boot.el:30154:(defalias 'vc-pull 'vc-update)
./loaddefs.el:29693:(defalias 'vc-update 'vc-pull)

and that vc-update is mentioned several times in autoload comments.

Re: Symbol's chain of function indirections contains a loop

[Andreas Schwab]

Lennart Borgman <lennart.borgman@gmail.com> writes:

> ./ldefs-boot.el:30154:(defalias 'vc-pull 'vc-update)

I have regenerated it.

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: Symbol's chain of function indirections contains a loop

[Eli Zaretskii]

> From: Lennart Borgman <lennart.borgman@gmail.com>
> Date: Sun, 30 Jan 2011 12:04:16 +0100
> Cc: Tassilo Horn <tassilo@member.fsf.org>, cschol2112@googlemail.com, emacs-devel@gnu.org
> 
> I have no idea of what is going on, but I notice these defaliases
> 
> ./ldefs-boot.el:30154:(defalias 'vc-pull 'vc-update)
> ./loaddefs.el:29693:(defalias 'vc-update 'vc-pull)

That was it!  Once Andreas regenerated ldefs-boot.el, Emacs now
bootstraps flawlessly on MS-Windows for me.

Does anyone else have problems bootstrapping on Windows?  If so,
please submit separate bug reports.

Thanks.

Re: Symbol's chain of function indirections contains a loop

[Christoph]

On 1/30/2011 11:21 AM, Eli Zaretskii wrote:

> That was it!  Once Andreas regenerated ldefs-boot.el, Emacs now
> bootstraps flawlessly on MS-Windows for me.
>
> Does anyone else have problems bootstrapping on Windows?  If so,
> please submit separate bug reports.

bzr r103035 builds fine on Windows.

Christoph

Re: Symbol's chain of function indirections contains a loop

[Lennart Borgman]

On Sun, Jan 30, 2011 at 7:21 PM, Eli Zaretskii <eliz@gnu.org> wrote:
>> From: Lennart Borgman <lennart.borgman@gmail.com>
>> Date: Sun, 30 Jan 2011 12:04:16 +0100
>> Cc: Tassilo Horn <tassilo@member.fsf.org>, cschol2112@googlemail.com, emacs-devel@gnu.org
>>
>> I have no idea of what is going on, but I notice these defaliases
>>
>> ./ldefs-boot.el:30154:(defalias 'vc-pull 'vc-update)
>> ./loaddefs.el:29693:(defalias 'vc-update 'vc-pull)
>
> That was it!  Once Andreas regenerated ldefs-boot.el, Emacs now
> bootstraps flawlessly on MS-Windows for me.


Thanks Eli and Andreas. Finally I was able to build Emacs again.

However the troubles we have discussed here were not all. I had
another one too which costed my some time. I am still not sure about
the problem and the solution to it (though I am quite sure it is w32
specific). Here it is:

For some time I have seen I have seen that the subprocess to running
the program diff in ediff hangs. I have to kill the program (I am
using Task Manager to do this). After that Emacs recieves the correct
info from diff and everything works as it should.

So it looks like Emacs is waiting for the subprocess to finish, doing
just the last thing whatever that might be, but something prevents it
from finishing.

I observed this with my patched Emacs and thought I would have a
closer look after a new checkout.

But then I saw that building Emacs also hanged. No CPU consumption. It
just hanged. It looked like it waited for a subprocess too.

So I searched the net and saw someone had problem with Google Desktop
and some compiler IDE (not Emacs). He did not find out what the
problem was but it went away when he uninstalled Google Desktop.

This was along my "suspicion line" where I suspect some update to my
antivirus software. Now I thought that Google Desktop probably used
the file change notification API:s too so it looked to me that they
may be doing some bad API call there.

I thought that there are more reason to suspect Google Desktop than
the antivirus software since the antivirus software is widely used
(and optimized not to disturb which probably mean they delay looking
at new and change files) on w32 while Google Desktop might not be that
much used (since w32 has a built in indexer). Beside that Google seems
to put less effort on w32. Rumors says they are not using w32 in
house. That of course means they might have less expertise on w32
API:s now.

So I disabled Google Desktop. And then I could build Emacs.

Though this does not mean I know where the problem really is. Could
there still be some problem with the w32 subprocess calls? Do we check
everything we can check after each call? And when it is finished?

Re: Symbol's chain of function indirections contains a loop

[Lennart Borgman]

On Sun, Jan 30, 2011 at 10:36 PM, Lennart Borgman
<lennart.borgman@gmail.com> wrote:
>
> However the troubles we have discussed here were not all. I had
> another one too which costed my some time. I am still not sure about
> the problem and the solution to it (though I am quite sure it is w32
> specific). Here it is:
>
> For some time I have seen I have seen that the subprocess to running
> the program diff in ediff hangs. I have to kill the program (I am
> using Task Manager to do this). After that Emacs recieves the correct
> info from diff and everything works as it should.
>
> So it looks like Emacs is waiting for the subprocess to finish, doing
> just the last thing whatever that might be, but something prevents it
> from finishing.
>
> I observed this with my patched Emacs and thought I would have a
> closer look after a new checkout.
>
> But then I saw that building Emacs also hanged. No CPU consumption. It
> just hanged. It looked like it waited for a subprocess too.
>
> So I searched the net and saw someone had problem with Google Desktop
> and some compiler IDE (not Emacs). He did not find out what the
> problem was but it went away when he uninstalled Google Desktop.
>
> This was along my "suspicion line" where I suspect some update to my
> antivirus software. Now I thought that Google Desktop probably used
> the file change notification API:s too so it looked to me that they
> may be doing some bad API call there.
>
> I thought that there are more reason to suspect Google Desktop than
> the antivirus software since the antivirus software is widely used
> (and optimized not to disturb which probably mean they delay looking
> at new and change files) on w32 while Google Desktop might not be that
> much used (since w32 has a built in indexer). Beside that Google seems
> to put less effort on w32. Rumors says they are not using w32 in
> house. That of course means they might have less expertise on w32
> API:s now.
>
> So I disabled Google Desktop. And then I could build Emacs.
>
> Though this does not mean I know where the problem really is. Could
> there still be some problem with the w32 subprocess calls? Do we check
> everything we can check after each call? And when it is finished?

I still get hangs when I build Emacs. Sometimes the subprocess running
Emacs hangs.

So it was probably not Google Desktop. Anyone have any idea of what it
might be? (I am disabling services one by one to try to find out, but
it is a bit time consuming. And errorprone since this an intermediate
error.)

Re: Symbol's chain of function indirections contains a loop

[Lennart Borgman]

On Tue, Feb 1, 2011 at 1:19 AM, Lennart Borgman
<lennart.borgman@gmail.com> wrote:
>>
>> So it looks like Emacs is waiting for the subprocess to finish, doing
>> just the last thing whatever that might be, but something prevents it
>> from finishing.

My last attempt to compile Emacs hanged with the last output

  make[1]: Leaving directory `C:/emacs-lp/bld/emacs/trunk/nt'
  Successfully remade target file `bootstrap'.

It is somewhere here:

    make --debug=b bootstrap 2>&1 | tee %trace-tmp%
    @findstr /m /c:"Error 2" %trace-tmp%
    @if %errorlevel% EQU 0 goto end
    cd %curr-wd%

    @if "%2"=="mkbootstrap" goto end
:mkinfo
    @echo.
    @echo.
    @echo **************************************************************
    @echo *** make info
    @echo **************************************************************

I have no idea about what it is waiting for. The latest days when I
have seen this it is usually a subprocess, but here I have no idea
what it could be. It looks like it is tee.exe that hangs here, but I
am not sure. (It used to work. Killing tee.exe stopped the build, but
it is a bit complicated by the fact that there are two tee processes.
I might have killed the wrong one or it is something else involved.)

Any hints? Anyone else seeing this kind of trouble on w32 lately?

Re: Symbol's chain of function indirections contains a loop

[Christoph]

On 1/31/2011 6:20 PM, Lennart Borgman wrote:

> It is somewhere here:
>
>      make --debug=b bootstrap 2>&1 | tee %trace-tmp%
>      @findstr /m /c:"Error 2" %trace-tmp%
>      @if %errorlevel% EQU 0 goto end
>      cd %curr-wd%
>
>      @if "%2"=="mkbootstrap" goto end
> :mkinfo
>      @echo.
>      @echo.
>      @echo **************************************************************
>      @echo *** make info
>      @echo **************************************************************

Is this one of your scripts?

> Any hints? Anyone else seeing this kind of trouble on w32 lately?

No, aside from the recent breakages due to gnulib I have had no problems 
building the trunk.

Christoph

Re: Symbol's chain of function indirections contains a loop

[Eli Zaretskii]

> Date: Mon, 31 Jan 2011 18:40:00 -0700
> From: Christoph <cschol2112@googlemail.com>
> CC: Eli Zaretskii <eliz@gnu.org>, tassilo@member.fsf.org, 
>  emacs-devel@gnu.org
> 
> No, aside from the recent breakages due to gnulib I have had no problems 
> building the trunk.

Same here.

Re: Symbol's chain of function indirections contains a loop

[Lennart Borgman]

On Tue, Feb 1, 2011 at 2:40 AM, Christoph <cschol2112@googlemail.com> wrote:
>>     @echo **************************************************************
>>     @echo *** make info
>>     @echo **************************************************************
>
> Is this one of your scripts?

Yes.

>> Any hints? Anyone else seeing this kind of trouble on w32 lately?
>
> No, aside from the recent breakages due to gnulib I have had no problems
> building the trunk.

Thanks.

Re: Symbol's chain of function indirections contains a loop

[Lennart Borgman]

On Tue, Feb 1, 2011 at 11:14 AM, Lennart Borgman
<lennart.borgman@gmail.com> wrote:
> On Tue, Feb 1, 2011 at 2:40 AM, Christoph <cschol2112@googlemail.com> wrote:
>>>     @echo **************************************************************
>>>     @echo *** make info
>>>     @echo **************************************************************
>>
>> Is this one of your scripts?
>
> Yes.
>
>>> Any hints? Anyone else seeing this kind of trouble on w32 lately?

I noticed there where some strange processes running ssl.exe from
strawberry perl so I tested to uninstall this perl and installing
activeperl instead. After this I tried building Emacs and now it
worked.

Though I am still not sure where the problem really is. I have also
disabled "Windows Search" and some other services. I might look deeper
into this if I get time. Hints still very much appreciated.

Re: Symbol's chain of function indirections contains a loop

[Lennart Borgman]

On Tue, Feb 1, 2011 at 1:11 PM, Lennart Borgman
<lennart.borgman@gmail.com> wrote:
> On Tue, Feb 1, 2011 at 11:14 AM, Lennart Borgman
> <lennart.borgman@gmail.com> wrote:
>> On Tue, Feb 1, 2011 at 2:40 AM, Christoph <cschol2112@googlemail.com> wrote:
>>>>     @echo **************************************************************
>>>>     @echo *** make info
>>>>     @echo **************************************************************
>>>
>>> Is this one of your scripts?
>>
>> Yes.
>>
>>>> Any hints? Anyone else seeing this kind of trouble on w32 lately?
>
> I noticed there where some strange processes running ssl.exe from
> strawberry perl so I tested to uninstall this perl and installing
> activeperl instead. After this I tried building Emacs and now it
> worked.
>
> Though I am still not sure where the problem really is. I have also
> disabled "Windows Search" and some other services. I might look deeper
> into this if I get time. Hints still very much appreciated.

I still see the problem when running ediff-buffers3. If I quickly do a
copy from C to A (i.e. "ca") and then a copy from C to B ("cb") then
Emacs hangs waiting for diff and diff3 to complete. If I kill diff.exe
everything works again.

If I do the above slowly there is no problem.