bug#31777: guile-2.2 FTCBFS for mipsel: In procedure load-thunk-from-memory: No such file or directory

bug#31777: guile-2.2 FTCBFS for mipsel: In procedure load-thunk-from-memory: No such file or directory

[Rob Browning]

[If possible, please preserve the 900203-forwarded address in replies]

We're currently unable to cross-build guile 2.2 on all of the debian
release architectures, which is important given that through make, etc.,
guile's now part of the core bootstrap set for new architectures.

You can see additional information here

  https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=900203

and I have been able to reproduce the failure on a debian buster
(testing) amd64 host by adding armhf as root:

  # dpkg --add-architecture armhf
  # apt build-dep -t buster guile-2.2
  # apt install -t buster \
    fakeroot \
    dpkg-dev \
    binutils:armhf \
    crossbuild-essential-armhf \
    gcc-arm-linux-gnueabihf \
    libc-dev:armhf
    libncurses5-dev:armhf

and then building the package as not-root:

  $ apt source guile-2.2=2.2.3+1-4
  $ cd guile-2.2-2.2.3+1
  $ DEB_BUILD_PROFILES="cross nocheck" \
    DEB_BUILD_OPTIONS="parallel=2 nocheck" \
    fakeroot dpkg-buildpackage -B --host-arch=armhf

...which fails with the same error reported in the debian bug:

  guild compile --target="arm-unknown-linux-gnueabihf" -Wunbound-variable -Wmacro-use-before-definition -Warity-mismatch -Wformat        \
    -L "/home/rlb/deb/guile/main/module" -L "/home/rlb/deb/guile/main/module"             \
    -L "/home/rlb/deb/guile/main/guile-readline"                  \
    --from=elisp -o "language/elisp/boot.go" "language/elisp/boot.el"
  Backtrace:
  In srfi/srfi-1.scm:
      640:9 19 (for-each #<procedure 55b6de6cca00 at scripts/compile.?> ?)
  In scripts/compile.scm:
     251:26 18 (_ _)
  In system/base/target.scm:
       57:6 17 (with-target _ _)
  In system/base/compile.scm:
     139:28 16 (compile-file "language/elisp/boot.el" #:output-file _ # ?)
  In system/base/language.scm:
     110:30 15 (default-environment _)
      62:11 14 (lookup-language elisp)
  In ice-9/boot-9.scm:
    2714:10 13 (_ (language elisp spec) _ _ #:ensure _)
    2982:16 12 (try-module-autoload _ _)
     2312:4 11 (save-module-excursion _)
    3002:22 10 (_)
  In unknown file:
             9 (primitive-load-path "language/elisp/spec" #<procedure ?>)
  In system/base/compile.scm:
      165:4  8 (compile-and-load _ #:from _ #:to _ #:env _ #:opts _ # _)
     235:18  7 (read-and-compile #<input: language/elisp/boot.el 13> # ?)
     183:32  6 (compile-fold (#<procedure compile-tree-il (expr env ?>) ?)
  In language/elisp/compile-tree-il.scm:
      805:5  5 (compile-tree-il (defmacro @ (module symbol) (#{`}# ?)) ?)
     705:11  4 (_ _ _)
  In system/base/compile.scm:
      255:6  3 (compile _ #:from _ #:to _ #:env _ #:opts _)
     183:32  2 (compile-fold _ #<tree-il (seq (call (@ (language elis?> ?)
  In language/bytecode/spec.scm:
      28:15  1 (bytecode->value #vu8(127 69 76 70 1 1 1 255 0 0 0 0 ?) ?)
  In unknown file:
             0 (load-thunk-from-memory #vu8(127 69 76 70 1 1 1 255 0 ?))

  ERROR: In procedure load-thunk-from-memory:
  In procedure load-thunk-from-memory: No such file or directory
  Makefile:2267: recipe for target 'language/elisp/boot.go' failed

Please let me know if I can help test anything further, and note that it
appears possible that the the architecture may be relevant, i.e. from
the debian bug:

Helmut Grohne <helmut@subdivi.de> writes:

> Initially, I thought the failure was 100% reproducible for any
> architecture. That doesn't seem to be the case. Let me try building
> guile-2.2 for most release architecture with sbuild:
>
> arm64: successful
> armel: ftcbfs, ftcbfs
> armhf: ftcbfs, ftcbfs
> mips: multiarch skew linux-libc-dev
> mips64el: successful
> mipsel: multiarch skew linux-libc-dev
> powerpc: ftcbfs, ftcbfs
> ppc64el: successful
> s390x: ftcbfs
>
> I ran each ftcbfs build twice to rule out the possibility of a random
> ftcbfs. So we have a non-random ftcbfs for some architectures. I'm a bit
> surprised about s390x here as it is the only failing 64bit architecture.

Thanks
-- 
Rob Browning
rlb @defaultvalue.org and @debian.org
GPG as of 2011-07-10 E6A9 DA3C C9FD 1FF8 C676 D2C4 C0F0 39E9 ED1B 597A
GPG as of 2002-11-03 14DD 432F AE39 534D B592 F9A0 25C8 D377 8C7E 73A4

bug#31777: guile-2.2 FTCBFS for mipsel: In procedure load-thunk-from-memory: No such file or directory

[Mark H Weaver]

Hi Rob,

Rob Browning <rlb@defaultvalue.org> writes:

> [If possible, please preserve the 900203-forwarded address in replies]
>
> We're currently unable to cross-build guile 2.2 on all of the debian
> release architectures, which is important given that through make, etc.,
> guile's now part of the core bootstrap set for new architectures.
>
> You can see additional information here
>
>   https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=900203

Thanks.  I was able to reproduce the problem by running:

  ./meta/build-env guild compile --target=arm-unknown-linux-gnueabihf \
    -Wunbound-variable -Wmacro-use-before-definition -Warity-mismatch \
    -Wformat -L module --from=elisp \
    -o module/language/elisp/ARM-boot.go \
    module/language/elisp/boot.el

from my built git checkout of the stable-2.2 branch.

>   ERROR: In procedure load-thunk-from-memory:
>   In procedure load-thunk-from-memory: No such file or directory
>   Makefile:2267: recipe for target 'language/elisp/boot.go' failed

The "No such file or directory" was bogus, caused by bad error handling
in 'load-thunk-from-memory', and fixed in the following commits:

  https://git.savannah.gnu.org/cgit/guile.git/commit/?h=stable-2.2&id=deef64a739d868ed5fda4e55e387528c1b6c196b
  https://git.savannah.gnu.org/cgit/guile.git/commit/?h=stable-2.2&id=9fbb36725634d05c3e46de7619e2f6019fbeb6fe

With those fixes, the error message becomes:

  In procedure load-thunk-from-memory: ELF file does not have native word size

While cross-compiling our elisp compiler, there were attempts to load
object code built for the target machine into the Guile process running
on the build machine.

I just pushed commit bff75635e5ed08de63b02c661cf65f2e1b28c995 to the
stable-2.2 branch, which together with the earlier commit
988aa29238fca862c7e2cb55f15762a69b4c16ce allows me to successfully
cross-compile language/elisp/boot.el using the above command.

Would you like to try cherry-picking these commits and see if they fix
the problem for you?

  https://git.savannah.gnu.org/cgit/guile.git/commit/?h=stable-2.2&id=988aa29238fca862c7e2cb55f15762a69b4c16ce
  https://git.savannah.gnu.org/cgit/guile.git/commit/?h=stable-2.2&id=bff75635e5ed08de63b02c661cf65f2e1b28c995

     Thanks,
       Mark

bug#31777: guile-2.2 FTCBFS for mipsel: In procedure load-thunk-from-memory: No such file or directory

[Rob Browning]

Mark H Weaver <mhw@netris.org> writes:

> Would you like to try cherry-picking these commits and see if they fix
> the problem for you?

Uploaded as 2.2.3+1-5.  Thanks for the help.

-- 
Rob Browning
rlb @defaultvalue.org and @debian.org
GPG as of 2011-07-10 E6A9 DA3C C9FD 1FF8 C676 D2C4 C0F0 39E9 ED1B 597A
GPG as of 2002-11-03 14DD 432F AE39 534D B592 F9A0 25C8 D377 8C7E 73A4

bug#31777: guile-2.2 FTCBFS for mipsel

[Vincent Legoll]

Hello,

Looks like this one should be closed.
Downstream is closed, marked as fixed.

Tell me if I did wrong.

-- 
Vincent Legoll