Re: [Bug-gnubg] gnubg 0.14.2 compiled successfully under HP-UX 11.0

[Mike Whitton]

 - tried regenerating the sgfl.c file and had the same outcome.

 - "/usr/include/sys/signal.h:207: parse error before `1'"
        + this was the error message produced during compilation of sgfl.c
           prior to adjusting it.

 - <stropts.h> is required for the ioctl command I_SETSIG
        + this is found on line 609 of 'sound.c'

 - ac_cv_func_mmap_fixed_mapped is set to no
        + HP-UX 11.0 does support mmap()
        + the static bearoff database setting causes problems for HP-UX
        + causes linking problems (unsatisfied symbols) during makehyper
           compilation.  looking for BearoffInitBuiltin() in eval.o
        + br1.c is only about 8kb when generated
        + all problems resolved when bearoff type switched to 'external'

 - the 'Manual example test' never runs successfully.
        + the 'gnubgtest' script runs and fails at this point.
        + it seems that the specific scenario needs to be setup manually.
        + in 'gnubgtest' the dice rolls seem to still be random.
        + using 'set rng manual' before starting a new game i was able to
           achieve the desired results.

thanks

Mike Whitton




On Mon, 29 Mar 2004, Joern Thyssen wrote:

> On Mon, Mar 29, 2004 at 12:42:21AM -0500, Mike Whitton wrote
> >
> > As an FYI, i have compiled gnubg 0.14.2 on HP-UX 11.0.  i compiled gnubg
> > using gcc 2.95.  i am playing right now and it works great.  congrats on
> > the fine software.  however, there were a few issues during compilation
> > that may help out your community:
> >
> >     -sgfl.c --> relocated '#include <stdlib.h>'
> >             + removed from if statement on line 1021 of sgfl.c
> >                and placed above other includes.
>
> sgfl.c is autogenerated by lex from sgf.l. The version generated by flex
> 2.5.4 seems to be unfit for HP-UX 11.0.
>
> Consider regenerating it:
>
> flex -osgfl.c -Psgf ./sgf.l
>
> >             + "/usr/include/sys/signal.h:207: parse error before `1'"
>
> I'm not sure what you mean by this?!
>
> >     -bearoff.h --> fTwoSided is missing
> >             + 'int fTwoSided;  /* one sided or two sided db? */'
> >             + added to the 'bearoffcontext' structure
>
> Fixed.
>
> This was a bug in makebearoff1.c which is the program that generated
> br1.c
>
> >     -sound.c --> missing header
> >             + added '#include <stropts.h>' to sound.c
>
> Which code or other include needs stropts.h?
>
> >     -during configure, bearoff database is incorrectly selected
> >             + attempts to generate and compile 'br1.c' by default
> >             + '--enable-bearoff=external' has to be specified
> >             + mmap() present
> >             + this may be attributed to my old automake (v1.4)
>
> Hmm, is ac_cv_func_mmap_fixed_mapped yes? Check in config.log
>
> >     -failure while running 'gnubgtest'
> >             + 'list game' test under Miscellaneous tests fails
> >             + after starting a new game 'list game' works fine
> >             + after starting a new match 'list game' seg faults
>
> I found a small bug in "list game", that could causes a seg fault.
>
> >             + also, the manual example game fails
> >                     - i tried running this manually
> >                     - do i need to specify the 'Match ID'?
>
> Can you be more specific about that fails?
>
> J?rn
>