bug#25961: Compiling new guix on old guix: 'waitpid' -1 failed unexpectedly: No child processes

bug#25961: Compiling new guix on old guix: 'waitpid' -1 failed unexpectedly: No child processes

[Christopher Allan Webber]

Hello!  I'm getting this error while trying to compile latest guix
master on one of my machines (fortunately, not my main one).  I don't
know why I'm hitting this error, since I'm not hitting it on my main
machine.  I'm upgrading from a few months back.

I spoke on irc and found that Ricardo had the same issue:

  <rekado> paroneayea: I also got this error trying to run “make”
  <rekado> paroneayea: it happens on my recording machine (x86_64) that hadn’t
           been updated in a while, but not on my laptop (also x86_64), which is
           updated regularly.

Here's what I do.  I start at the commit I was at
(c2662820f359be19262cdd5d564e6a0dddc43281) and run "guix environment
guix".  Then I do a "git pull && make clean && make" and I get:

  ;;; Failed to autoload cairo-image-surface-create in (cairo):
  ;;; ERROR: missing interface for module (cairo)
  ;;; Failed to autoload cairo-image-surface-create in (cairo):
  ;;; ERROR: missing interface for module (cairo)
  ;;; Failed to autoload rsvg-handle-new-from-file in (rsvg):
  ;;; ERROR: missing interface for module (rsvg)
  ;;; Failed to autoload cairo-image-surface-create in (cairo):
  ;;; ERROR: missing interface for module (cairo)
  ;;; Failed to autoload rsvg-handle-new-from-file in (rsvg):
  ;;; ERROR: missing interface for module (rsvg)
  ;;; Failed to autoload cairo-image-surface-create in (cairo):
  ;;; ERROR: missing interface for module (cairo)
    LOAD     (gnu build vm)
    LOAD     (gnu tests)
    LOAD     (gnu tests base)
    LOAD     (gnu tests nfs)
    LOAD     (gnu tests install)
  warning: 'waitpid' -1 failed unexpectedly: No child processes
  Backtrace:
  In ice-9/boot-9.scm:
  3088: 19 [#<procedure 4226c30 at ice-9/boot-9.scm:3069:17 ()>]
  In unknown file:
     ?: 18 [primitive-load-path "gnu/tests/install" ...]
  In ice-9/eval.scm:
   453: 17 [eval # ()]
   411: 16 [eval # #]
   387: 15 [eval # #]
   373: 14 [run-install # #]
   387: 13 [eval # #]
   387: 12 [eval # #]
   387: 11 [eval # #]
   411: 10 [eval # #]
  In srfi/srfi-1.scm:
   573: 9 [map #<procedure 360a6c0 at ice-9/eval.scm:416:20 (a)> (# # # # ...)]
  In ice-9/eval.scm:
   387: 8 [eval # #]
   411: 7 [eval # #]
   411: 6 [eval # #]
   387: 5 [eval # #]
  In unknown file:
     ?: 4 [force #<promise #<procedure 5bbf150 at ice-9/eval.scm:416:20 ()>>]
  In ice-9/eval.scm:
   411: 3 [eval # #]
   411: 2 [eval # #]
  In ice-9/popen.scm:
    98: 1 [close-process #<closed: file 0> 28701]
  In unknown file:
     ?: 0 [waitpid 28701 #<undefined>]

  ERROR: In procedure waitpid:
  ERROR: In procedure waitpid: No child processes
  make[2]: *** [Makefile:4929: make-go] Error 1
  make[2]: Leaving directory '/home/cwebber/src/guix'
  make[1]: *** [Makefile:4072: all-recursive] Error 1
  make[1]: Leaving directory '/home/cwebber/src/guix'
  make: *** [Makefile:2728: all] Error 2

This is on commit 21abf092a49f0ce80cbfff5cccabb7dbf53abf96

If instead I switch back to c2662820f359be19262cdd5d564e6a0dddc43281 I
can compile Guix again.

I did a git-bisect to find what commit was breaking things.  According
to git-bisect, the perpetrator is:

  63302a4e55241a41eab4c21d7af9fbd0d5817459 is the first bad commit
  commit 63302a4e55241a41eab4c21d7af9fbd0d5817459
  Author: Ludovic Courtès <ludo@gnu.org>
  Date:   Mon Feb 6 23:47:09 2017 +0100

      Add (gnu build shepherd).

      * gnu/build/shepherd.scm: New file.
      * gnu/local.mk (GNU_SYSTEM_MODULES): Add it.

  :040000 040000 7a07b30ebec719aca2fba4c2776f7152e669a1ee 63eb6bc3fa184fc42c3d469d594fc056f16b7892 M	gnu
  bisect run success

I've confirmed this by double-checking that I can build the commit
before but not this commit.

I don't really know why this fails, but it looks like it's somehow
triggering a remote process in some way related to tests and
shepherd... which is confusing, because I don't think this happening
during tests.

 - Chris

bug#25961: Compiling new guix on old guix: 'waitpid' -1 failed unexpectedly: No child processes

[Marius Bakke]

[-- Attachment #1: Type: text/plain, Size: 4049 bytes --]

Hi!

Christopher Allan Webber <cwebber@dustycloud.org> writes:

> Hello!  I'm getting this error while trying to compile latest guix
> master on one of my machines (fortunately, not my main one).  I don't
> know why I'm hitting this error, since I'm not hitting it on my main
> machine.  I'm upgrading from a few months back.
>
> I spoke on irc and found that Ricardo had the same issue:
>
>   <rekado> paroneayea: I also got this error trying to run “make”
>   <rekado> paroneayea: it happens on my recording machine (x86_64) that hadn’t
>            been updated in a while, but not on my laptop (also x86_64), which is
>            updated regularly.
>
> Here's what I do.  I start at the commit I was at
> (c2662820f359be19262cdd5d564e6a0dddc43281) and run "guix environment
> guix".  Then I do a "git pull && make clean && make" and I get:
>
>   ;;; Failed to autoload cairo-image-surface-create in (cairo):
>   ;;; ERROR: missing interface for module (cairo)
>   ;;; Failed to autoload cairo-image-surface-create in (cairo):
>   ;;; ERROR: missing interface for module (cairo)
>   ;;; Failed to autoload rsvg-handle-new-from-file in (rsvg):
>   ;;; ERROR: missing interface for module (rsvg)
>   ;;; Failed to autoload cairo-image-surface-create in (cairo):
>   ;;; ERROR: missing interface for module (cairo)
>   ;;; Failed to autoload rsvg-handle-new-from-file in (rsvg):
>   ;;; ERROR: missing interface for module (rsvg)
>   ;;; Failed to autoload cairo-image-surface-create in (cairo):
>   ;;; ERROR: missing interface for module (cairo)
>     LOAD     (gnu build vm)
>     LOAD     (gnu tests)
>     LOAD     (gnu tests base)
>     LOAD     (gnu tests nfs)
>     LOAD     (gnu tests install)
>   warning: 'waitpid' -1 failed unexpectedly: No child processes
>   Backtrace:
>   In ice-9/boot-9.scm:
>   3088: 19 [#<procedure 4226c30 at ice-9/boot-9.scm:3069:17 ()>]
>   In unknown file:
>      ?: 18 [primitive-load-path "gnu/tests/install" ...]
>   In ice-9/eval.scm:
>    453: 17 [eval # ()]
>    411: 16 [eval # #]
>    387: 15 [eval # #]
>    373: 14 [run-install # #]
>    387: 13 [eval # #]
>    387: 12 [eval # #]
>    387: 11 [eval # #]
>    411: 10 [eval # #]
>   In srfi/srfi-1.scm:
>    573: 9 [map #<procedure 360a6c0 at ice-9/eval.scm:416:20 (a)> (# # # # ...)]
>   In ice-9/eval.scm:
>    387: 8 [eval # #]
>    411: 7 [eval # #]
>    411: 6 [eval # #]
>    387: 5 [eval # #]
>   In unknown file:
>      ?: 4 [force #<promise #<procedure 5bbf150 at ice-9/eval.scm:416:20 ()>>]
>   In ice-9/eval.scm:
>    411: 3 [eval # #]
>    411: 2 [eval # #]
>   In ice-9/popen.scm:
>     98: 1 [close-process #<closed: file 0> 28701]
>   In unknown file:
>      ?: 0 [waitpid 28701 #<undefined>]
>
>   ERROR: In procedure waitpid:
>   ERROR: In procedure waitpid: No child processes
>   make[2]: *** [Makefile:4929: make-go] Error 1
>   make[2]: Leaving directory '/home/cwebber/src/guix'
>   make[1]: *** [Makefile:4072: all-recursive] Error 1
>   make[1]: Leaving directory '/home/cwebber/src/guix'
>   make: *** [Makefile:2728: all] Error 2
>
> This is on commit 21abf092a49f0ce80cbfff5cccabb7dbf53abf96
>
> If instead I switch back to c2662820f359be19262cdd5d564e6a0dddc43281 I
> can compile Guix again.
>
> I did a git-bisect to find what commit was breaking things.  According
> to git-bisect, the perpetrator is:
>
>   63302a4e55241a41eab4c21d7af9fbd0d5817459 is the first bad commit
>   commit 63302a4e55241a41eab4c21d7af9fbd0d5817459
>   Author: Ludovic Courtès <ludo@gnu.org>
>   Date:   Mon Feb 6 23:47:09 2017 +0100
>
>       Add (gnu build shepherd).
>
>       * gnu/build/shepherd.scm: New file.
>       * gnu/local.mk (GNU_SYSTEM_MODULES): Add it.
>
>   :040000 040000 7a07b30ebec719aca2fba4c2776f7152e669a1ee 63eb6bc3fa184fc42c3d469d594fc056f16b7892 M	gnu
>   bisect run success
>
> I've confirmed this by double-checking that I can build the commit
> before but not this commit.

Which kernel version is this?

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 487 bytes --]

bug#25961: Compiling new guix on old guix: 'waitpid' -1 failed unexpectedly: No child processes

[Christopher Allan Webber]

Marius Bakke <mbakke@fastmail.com> writes:

> Which kernel version is this?

Linux-Libre 4.8.1

And!  As it turns out, I went to the last revision before I was hitting
compilation isues (c90db25f4cf1f98f3f4f3af38d175a14ffb8c32a), did a make
on the repo, and from there I did a "guix system reconfigure".  I
restarted, and after that I was able to build the repo on git master. 
If it's because of the kernel that things work now, it did upgrade the
kernel... the kernel is now at 4.9.8.

I wonder why this happened?

bug#25961: Compiling new guix on old guix: 'waitpid' -1 failed unexpectedly: No child processes

[Marius Bakke]

This was likely due to ABI breakage, which is caught and reported
automatically in recent versions of Guix.