m68k UnoApiTarget/udkapi.rdb build fails ("Bad input")

classic Classic list List threaded Threaded
7 messages Options
Aaron M. Ucko Aaron M. Ucko
Reply | Threaded
Open this post in threaded view
|

m68k UnoApiTarget/udkapi.rdb build fails ("Bad input")

At least on Debian's autobuilders, builds of LibreOffice for m68k have
been failing since 6.0.0 rc2 with a "Bad input" error when trying to build
UnoApiTarget/udkapi.rdb, per the below excerpt from
https://buildd.debian.org/status/fetch.php?pkg=libreoffice&arch=m68k&ver=1%3A6.0.0-1&stamp=1517451030&raw=0:

  [build DBc] udkapi
mkdir -p /<<PKGBUILDDIR>>/workdir/UnoApiTarget/  && RESPONSEFILE=/tmp/gbuild.RTQBLH && LD_LIBRARY_PATH=${LD_LIBRARY_PATH:+$LD_LIBRARY_PATH:}"/<<PKGBUILDDIR>>/instdir/program:/<<PKGBUILDDIR>>/instdir/program"   /<<PKGBUILDDIR>>/workdir/LinkTarget/Executable/unoidl-write  /<<PKGBUILDDIR>>/udkapi @${RESPONSEFILE} /<<PKGBUILDDIR>>/workdir/UnoApiTarget/udkapi.rdb && rm -f ${RESPONSEFILE}   && LD_LIBRARY_PATH=${LD_LIBRARY_PATH:+$LD_LIBRARY_PATH:}"/<<PKGBUILDDIR>>/instdir/program:/<<PKGBUILDDIR>>/instdir/program"   /<<PKGBUILDDIR>>/instdir/sdk/bin/unoidl-check /<<PKGBUILDDIR>>/udkapi/type_reference/udkapi.idl --  /<<PKGBUILDDIR>>/workdir/UnoApiTarget/udkapi.rdb
  [build MOD] codemaker
  Bad input <file:///<<PKGBUILDDIR>>/udkapi/com/sun/star/reflection/TypeDescriptionSearchDepth.idl>: cannot parse line 37: "out-of-range enum com.sun.star.reflection.TypeDescriptionSearchDepth member INFINITE value 9223372036854775807"
  /<<PKGBUILDDIR>>/solenv/gbuild/UnoApiTarget.mk:45: recipe for target '/<<PKGBUILDDIR>>/workdir/UnoApiTarget/udkapi.rdb' failed
  make[2]: *** [/<<PKGBUILDDIR>>/workdir/UnoApiTarget/udkapi.rdb] Error 1
  make[2]: *** Waiting for unfinished jobs....

Could somebody please take a look?

Thanks!

FTR, this report corresponds to
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=889605 and
https://bugs.documentfoundation.org/show_bug.cgi?id=115450
(closed as out of scope; sorry for the noise there).

--
Aaron M. Ucko, KB1CJC (amu at alum.mit.edu, ucko at debian.org)
http://www.mit.edu/~amu/ | http://stuff.mit.edu/cgi/finger/?amu@...
_______________________________________________
LibreOffice mailing list
[hidden email]
https://lists.freedesktop.org/mailman/listinfo/libreoffice
Caolán McNamara Caolán McNamara
Reply | Threaded
Open this post in threaded view
|

Re: m68k UnoApiTarget/udkapi.rdb build fails ("Bad input")

On Mon, 2018-02-05 at 21:41 -0500, Aaron M. Ucko wrote:
> builds of LibreOffice for m68k
...
> <file:///<<PKGBUILDDIR>>/udkapi/com/sun/star/reflection/TypeDescripti
> onSearchDepth.idl>: cannot parse line 37: "out-of-range enum
> com.sun.star.reflection.TypeDescriptionSearchDepth member INFINITE
> value 9223372036854775807"
...
> Could somebody please take a look?

To be honest I added m68k support back in the day more for comedic
effect than anything else so unless rene has any ideas I think you're
probably on your own on debugging that.
_______________________________________________
LibreOffice mailing list
[hidden email]
https://lists.freedesktop.org/mailman/listinfo/libreoffice
Rene Engelhard Rene Engelhard
Reply | Threaded
Open this post in threaded view
|

Re: m68k UnoApiTarget/udkapi.rdb build fails ("Bad input")

Hi,

On Tue, Feb 06, 2018 at 09:02:58PM +0000, Caolán McNamara wrote:

> On Mon, 2018-02-05 at 21:41 -0500, Aaron M. Ucko wrote:
> > builds of LibreOffice for m68k
> ...
> > <file:///<<PKGBUILDDIR>>/udkapi/com/sun/star/reflection/TypeDescripti
> > onSearchDepth.idl>: cannot parse line 37: "out-of-range enum
> > com.sun.star.reflection.TypeDescriptionSearchDepth member INFINITE
> > value 9223372036854775807"
> ...
> > Could somebody please take a look?
>
> To be honest I added m68k support back in the day more for comedic
> effect than anything else so unless rene has any ideas I think you're
> probably on your own on debugging that.

Don't have (and yes, I saw the build failure and your bug report).

Quickly looked at the file, but tbh don't see why it should fail only on
m68k (no idea about arch-specifics which can cause that, though).

The latest round of m68k fixes (-mmlong-jump-table-offsets..) was done by
John Paul Adrian Glaubitz, though, maybe he has an idea.

Otherwise I am in the same boat as Caolan :)

Regards,

Rene
_______________________________________________
LibreOffice mailing list
[hidden email]
https://lists.freedesktop.org/mailman/listinfo/libreoffice
sberg sberg
Reply | Threaded
Open this post in threaded view
|

Re: m68k UnoApiTarget/udkapi.rdb build fails ("Bad input")

In reply to this post by Aaron M. Ucko
On 06.02.2018 03:41, Aaron M. Ucko wrote:
>    Bad input <file:///<<PKGBUILDDIR>>/udkapi/com/sun/star/reflection/TypeDescriptionSearchDepth.idl>: cannot parse line 37: "out-of-range enum com.sun.star.reflection.TypeDescriptionSearchDepth member INFINITE value 9223372036854775807"

udkapi/com/sun/star/reflection/TypeDescriptionSearchDepth.idl:37 is

     INFINITE = -1,

where "1" is lexed as TOK_INTEGER with sal_uInt64 value 1 in
unoidl/source/sourceprovider-scanner.l (see nonZeroIntegerLiteral), and
the negation is parsed as a unaryExpr in
unoidl/source/sourceprovider-parser.y, converting the
unoidl::detail::SourceProviderExpr::Uint(1) to
unoidl::detail::SourceProviderExpr::Int(-1).  Then the whole enum member
declaration is parsed as enumMember (second branch, with explicit
initializer expr), but shouldn't cause the out-of-range check in the
unoidl::detail::SourceProviderExpr::TYPE_INT case to fire.

Odd, esp. as 9223372036854775807 is 0x7FFFFFFFFFFFFFFF, not
0xFFFFFFFFFFFFFFFF (aka -1). Smells like an issue with the compilation
toolchain.
_______________________________________________
LibreOffice mailing list
[hidden email]
https://lists.freedesktop.org/mailman/listinfo/libreoffice
John Paul Adrian Glaubitz John Paul Adrian Glaubitz
Reply | Threaded
Open this post in threaded view
|

Re: m68k UnoApiTarget/udkapi.rdb build fails ("Bad input")

In reply to this post by Rene Engelhard
Hi!

On 02/06/2018 10:29 PM, Rene Engelhard wrote:
> Don't have (and yes, I saw the build failure and your bug report).
>
> Quickly looked at the file, but tbh don't see why it should fail only on
> m68k (no idea about arch-specifics which can cause that, though).
>
> The latest round of m68k fixes (-mmlong-jump-table-offsets..) was done by
> John Paul Adrian Glaubitz, though, maybe he has an idea.

Yes, I will have a look at it at some point.

Please bear with me as I am very busy with many projects I am hacking on
to fix such bugs. We are still actively maintaining m68k in Debian and
these issues are being looked at.

It sometimes takes a bit time, so please do not put any pressure on us
in the sense that you are going to remove the m68k code altogether.

Thanks,
Adrian

--
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - [hidden email]
`. `'   Freie Universitaet Berlin - [hidden email]
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913
_______________________________________________
LibreOffice mailing list
[hidden email]
https://lists.freedesktop.org/mailman/listinfo/libreoffice
Aaron M. Ucko Aaron M. Ucko
Reply | Threaded
Open this post in threaded view
|

Re: m68k UnoApiTarget/udkapi.rdb build fails ("Bad input")

In reply to this post by sberg
Stephan Bergmann <[hidden email]> writes:

> Odd, esp. as 9223372036854775807 is 0x7FFFFFFFFFFFFFFF, not
> 0xFFFFFFFFFFFFFFFF (aka -1). Smells like an issue with the compilation
> toolchain.

Entirely possible, as there was in fact an intervening GCC upgrade, with
the following ChangeLog stanza:

gcc-7 (7.2.0-19) unstable; urgency=medium

  * Update to SVN 20180107 (r256317) from the gcc-7-branch.
    - Fix PR libstdc++/82522, PR debug/83550, PR middle-end/83608,
      PR middle-end/83609, PR middle-end/83623, PR c++/83553, PR lto/82027,
      PR c/83448, PR rtl-optimization/80747, PR rtl-optimization/83512,
      PR tree-optimization/83523, PR tree-optimization/83521, PR ipa/82801,
      PR ipa/83346, PR target/83467 (x86), PR tree-optimization/82726,
      PR tree-optimization/70754, PR target/83387 (PPC), PR c++/83556,
      PR c++/83116, PR Fortran/83679, PR fortran/83650, PR Fortran/83548,
      PR libgfortran/83649, PR libgfortran/83613, PR libgfortran/81937,
      PR libgfortran/78549, PR libstdc++/83600, PR libstdc++/83598,
      PR libstdc++/83279, PR libstdc++/83626, PR libstdc++/83626,
      PR preprocessor/83492.
  * debian/rules2: Fix typo for N32 conditions (YunQiang Su). Closes: #886316.
  * More libffi mips r6 updates (YunQiang Su). Addresses: #886201.
  * libgo: Backport upstream sparc64 fix (James Clark). Closes: #884642.
  * Default to PIE on the hurd (Samuel Thibault). Closes: #885056.
  * Use internal libunwind for ia64 cross-builds. Closes: #885931.
  * Strip -z,defs from linker options for internal libunwind (James Clarke).
    Closes: #885937.
  * Fix rtlibs stage build with debhelper 10.9.1 (Helmut Grohne).
    Closes: #879054.

 -- Matthias Klose <[hidden email]>  Sun, 07 Jan 2018 08:55:48 +0100

Comparing the "Build environment" sections of [1] (good) and [2] (bad)
also indicates some other upgrades, though GCC is my prime suspect if
this is in fact a toolchain regression.
- binutils 2.29.1-2 -> 2.29.1-3, featuring a fix for PR ld/22626.
- glibc 2.25-5 -> 2.26-3.
- linux-libc-dev (kernel headers) 4.14.2-1 -> 4.14.12-2.
- Upgrades to various libraries and utilities outside the central
  toolchain.

Thanks for the analysis, and sorry for the noise if this turns out to be
a toolchain bug.

[1] https://buildd.debian.org/status/fetch.php?pkg=libreoffice&arch=m68k&ver=1%3A6.0.0%7Erc1-1&stamp=1514303458&raw=0
[2] https://buildd.debian.org/status/fetch.php?pkg=libreoffice&arch=m68k&ver=1%3A6.0.0%7Erc2-1&stamp=1516003390&raw=0

--
Aaron M. Ucko, KB1CJC (amu at alum.mit.edu, ucko at debian.org)
http://www.mit.edu/~amu/ | http://stuff.mit.edu/cgi/finger/?amu@...
_______________________________________________
LibreOffice mailing list
[hidden email]
https://lists.freedesktop.org/mailman/listinfo/libreoffice
Aaron M. Ucko Aaron M. Ucko
Reply | Threaded
Open this post in threaded view
|

Re: m68k UnoApiTarget/udkapi.rdb build fails ("Bad input")

"Aaron M. Ucko" <[hidden email]> writes:

> Thanks for the analysis, and sorry for the noise if this turns out to be
> a toolchain bug.

For the record, the error indeed turned out to lie in the build
environment (specifically, qemu):

https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=889605#21

--
Aaron M. Ucko, KB1CJC (amu at alum.mit.edu, ucko at debian.org)
http://www.mit.edu/~amu/ | http://stuff.mit.edu/cgi/finger/?amu@...
_______________________________________________
LibreOffice mailing list
[hidden email]
https://lists.freedesktop.org/mailman/listinfo/libreoffice