#2 Fixed: Gnome version will not build

open-fixed
nobody
None
9
2003-04-01
2003-03-30
Kyle Ferrio
No

I tried building the Gnome version from the source rpm.
No joy. The create_data perl script complains about a
verb which does not make sense, and then exits:

Mmm, this record is wrong ...
[teir-8>dos_part=teido|tinto<1FUTs:ti-e-re;1PASs:ti-e-ra;
1PASsB:ti-e-se;1PRE:ti-o;1PREs:ti-a;2FUTs:ti-e-res;2IMP:ti
-e;2PASs:ti-e-ras;2PASsB:ti-e-ses;2PRE:ti-es;2PREs:ti-as;
3FUTs:ti-e-re;3IMP:ti-a;3PAS:ti-;3PASs:ti-e-ra;3PASsB:ti-
e-se;3PRE:ti-e;3PREs:ti-a;4FUTs:ti-�re-mos;4IMP:ti-a-
mos;4PASs:ti-�ra-mos;4PASsB:ti-�se-mos;4PREs:ti-a-mos;
5FUTs:ti-e-reis;5PASs:ti-e-rais;5PASsB:ti-e-seis;5PREs:ti
-�s;6FUTs:ti-e-ren;6IMP:ti-an;6PAS:ti-e-ron;6PASs:ti-e-
ran;6PASsB:ti-e-sen;6PRE:ti-en;6PREs:ti-an;INF:te-ir;gerundio:ti-en-do;participio:tinto;participiocomp:teido;]
make[2]: *** [compjugador.gdbm] Error 25

Oddly, the text "teir" does not seem to appear in any
of the verb textfiles. I am going to go back and try
to build from the tarball, in case its different from
the tarball in the src rpm.

Mandrake 9.1
gcc 3.2.2
gdbm2
./configure --with-gnome=/usr

Discussion

  • Kyle Ferrio
    Kyle Ferrio
    2003-03-30

    Logged In: YES
    user_id=745416

    I tried building from the plain 0.3 tarball. The
    create_data script still dies in the same way. But this
    time I actually found the 'teir' verb in ver.con.txt. (I
    guess my text-viewer was having some problems with escapes.)
    Bottom line: it still does not build. The perl script
    appears to be unable to parse its own data. I'm running
    perl 5.8.0.

     
  • Kyle Ferrio
    Kyle Ferrio
    2003-03-30

    • priority: 5 --> 9
     
  • Kyle Ferrio
    Kyle Ferrio
    2003-03-31

    Logged In: YES
    user_id=745416

    Update: I still don't have a build, but I'm learning some
    things. The focus here is on the perl script create_data
    which parses the ver*.txt files to create the database.

    (1) During the parsing of ver.con.txt, the script bombs out
    on the infinitive te~nir. (In the file, it's an en-neh, but
    I'm writing it as a digraph in case SourceForge can't render
    it.) Commenting that verb out at least allows the
    processing of ver.con.txt to finish. Note that te~nir is
    not the first occurance of en-neh in that file. For
    example, so~nar.

    (2) With te~nir commented out, processing procedes to
    ver.fam.txt. Here, perl complains three times about
    malformed UTF-8 characters before bombing. Interestingly,
    the three lines it complains about are the first three lines
    containing en-neh.

    (3) I'm clueless about UTF-8. My guess is that my system is
    somehow mis-configured. Maybe I need to recompile perl with
    better support for internationalization. Any pointers will
    be greatly appreciated.

    (4) Mandrake 9.1
    Perl 5.8.0
    gcc 3.2.2
    gdbm2

     
  • Kyle Ferrio
    Kyle Ferrio
    2003-03-31

    Logged In: YES
    user_id=745416

    Yep. Commenting out the ~n verbs helps. Perl also
    complains about the arg''uir, so it seems that my build of
    perl does not like this. I feel like I'm getting close...

     
  • Kyle Ferrio
    Kyle Ferrio
    2003-04-01

    • summary: Gnome version will not build --> Fixed: Gnome version will not build
    • status: open --> open-fixed
     
  • Kyle Ferrio
    Kyle Ferrio
    2003-04-01

    Logged In: YES
    user_id=745416

    Warning! Be sure your editor supports the extended Latin
    character set before editing create_data. NEdit does.
    Emacs is probably fine, of course. My installed vi(m) does
    not, and that was a source of some wasted time! Ditto for
    the KDE gui editors I tried.

    Early on in create_data.pl, perl 5.8 (and perhaps as far
    back as 5.6?) needs to see this:

    require Encode;

    Then, after each line is read in, you'll need

    Encode::_utf8_off($_);

    Presumably man perlutf is of some help here...

    By the way, the instructions in the INSTALL file are a
    little out of sync. As built, the test program will fail
    unless you give it the full path of the database, which is
    built in the data directory.

    Finally, this onty resolves the build problem. It still
    segfaults on many verbs, including but not limited to

    leer, creer, venir, decir, seguir.

    Some verbs are just plain wrong:
    estar gives yo esto
    ser gives yo so
    dar gives yo do... you get the idea.

    Finally, dar and ser are the only two verbsI've noticed with
    conjugations for vosotros. Most of the verbs I've tried
    just have a blank there.