Re: [Gcl-devel] running program compiled by GCL via Wine

[Donald Winiecki]

Apologies. I have not run MAKE INSTALL in quite awhile! Will run
through that and see if I can repeat and trace the issue.

BTW, I don't use Wine (that I know of!).

Best,

_don



On Thu, Oct 17, 2013 at 8:26 AM, Camm Maguire <address@hidden> wrote:
> Greetings!
>
> Robert Dodier <address@hidden> writes:
>
>> Camm, thanks for the info about Wine & Maxima.
>>
>> I set up the binfmt stuff to recognize .exe files and it
>> seems to work as intended (.exe is indeed executed
>> via Wine) but I get the same error about msys.
>>
>> I don't understand the logic of detect_wine -- I don't
>> see how it could succeed. Maybe you can help me
>> understand it.
>>
>> I tried to execute just GCL itself  by installing GCL 2.6.9
>> ANSI from the download page for the GCL project.
>> The installer runs OK (some warnings) but saved_ansi_gcl.exe
>> fails with the same error which originates in detect_wine.
>> Here's how it reads when I try to execute saved_ansi_gcl.exe:
>>
>> Can't recognize 'c:/_cvs/gcl/lib/gcl-2.6.9/unixport/msys /tmp/ out8
>> tmp8 log8' as an internal or external command, or batch script.
>>
>> I see that the installer installed msys.exe in the folder
>> GCL-2.6.9-ANSI/unixport, but saved_ansi_gcl.exe tries
>> to execute c:/_cvs/gcl/lib/gcl-2.6.9/unixport/msys instead.
>>
>> Looks like the immediate problem is that *SYSTEM-DIRECTORY*
>> is not assigned a local path name but gets a hardcoded value
>> (something set at build time, I guess) instead.
>> Reading o/main.c, I see that GCL will accept a command
>> line argument for *SYSTEM-DIRECTORY* if initflag == 0,
>> which is not the case for saved images but is for raw images,
>> according to a comment in main.c.
>>
>
> Thanks so much for pointing this out!  All my experience, mostly
> successful, with running under wine comes from running in the build
> directory.  You indeed seem to have discovered a problem with our
> windows installer.  'make install' is supposed to run the following to
> setup the system directory paths to their correct final values:
>
> From the top level makefile:
>
>      echo '(reset-sys-paths "$(INSTALL_LIB_DIR)/")(si::save-system 
> "$(FLISP)$(EXE)")' | ./temp$(EXE) && \
>
> Don, can you see any reason why this is not happening?
>
> In any case, alas, the detect_wine function is run too early to change
> the si::*system-directory*.  I suppose I could have it read an
> environment variable by default, or maybe move it to the first use of
> 'system', which might provide more working room.  Suggestions most
> welcome.  In any case, if there is a chance you might build gcl from
> source according to README.wine and run from the build directory, I'd
> appreciate any feedback on how it goes for you.
>
>> So ... I can get a little farther by executing raw_ansi_gcl.exe.
>>
>> wine ./drive_c/ProgramFiles/GCL-2.6.9-ANSI/unixport/raw_ansi_gcl.exe
>> `pwd`/drive_c/ProgramFiles/GCL-2.6.9-ANSI/unixport/
>> GCL (GNU Common Lisp)  April 1994  131071 pages
>> Building symbol table for
>> Z:/home/robert/.wine/drive_c/ProgramFiles/GCL-2.6.9-ANSI/unixport/raw_ansi_gcl.exe
>> ..
>> loading 
>> /home/robert/.wine/drive_c/ProgramFiles/GCL-2.6.9-ANSI/unixport/../lsp/gcl_export.lsp
>>
>> Then it just sits there until I hit ctrl-C.
>>
>
> The raw image will be of no use to you unless you have access to the
> build directory.
>
>> strace shows that it seems to be repeatedly reading from
>> z:/tmp/out8 which has the content
>>
>> ar x 
>> /home/robert/.wine/drive_c/ProgramFiles/GCL-2.6.9-ANSI/unixport/libansi_gcl.a
>> gcl_defmacro.o
>>
>
> msys is just a little helper to run external system commands
> synchronously.  My experience with wine was that the native call
> returned before the child finished, leading to corrupt gcc .o output,
> for example.  From the above lines, yo don't have libansi_gcl.a (I don't
> think) installed, so ar cannot produce the needed file to run the
> initialization in the raw image.
>
>> I guess this is related to scheme implemented by msys --
>> not sure how this is supposed to play out.
>>
>> Thanks for any light you can shed on this.
>>
>
> Hope this is clear enough -- if not, please ask more questions.  I will
> think about a detect_wine system-directory escape, and hopefully if you
> can you might be able to test from a build directory.
>
> Take care,
> --
> Camm Maguire                                        address@hidden
> ==========================================================================
> "The earth is but one country, and mankind its citizens."  --  Baha'u'llah