[colm-users] Hardcoded word_t

[Konstantin Podsvirov]

Hello again!

11.05.2018, 20:11, "Adrian Thurston" <thurston at colm.net>:
> We'll get this cleaned up eventually. There is nothing about the
> language design that demands certain sizes for things. It's just a
> matter of me doing most of my research on a certain platform. It's
> important to note that colm is still a research project. Ragel 7 depends
> on it now, but that is still also a development release.
>
> Note I've got some changes to colm to push that fix certain issues that
> arise on platforms other than the one I work on mostly (amd64 ubuntu
> 16.04). Coming soon.

I have solution :-)

Today after latest changes in `master` branch I can compile `coml` with MinGW-w64 64bit toolchains.

Your can try changes from `word` branch on my server:

http://git.podsvirov.pro/?p=dad/fix/colm.git;a=shortlog;h=refs/heads/word

All changes in one last commit:

The type word_t declaration

If sizeof(unsigned long) == sizeof(void *) then word_t is unsigned long.
Else if sizeof(unsigned long long) == sizeof(void *) then word_t is unsigned long long.

Compilation checked with MSYS2 toolchains via Autotools project and
with MinGW-w64 toolchains via CMake project.

http://git.podsvirov.pro/?p=dad/fix/colm.git;a=commitdiff;h=63224902a50c68b3ce4e38827199bae4d50746f3

Also available via git:

git clone -b word git://podsvirov.pro/dad/fix/colm.git

Adrian, please review my changes and apply if posible.

> On 2018-05-11 05:05, Konstantin Podsvirov wrote:
>>  03.05.2018, 21:13, "Konstantin Podsvirov" <konstantin at podsvirov.pro>:
>>>>>  20:12, 3 May 2018 г., Adrian Thurston <thurston at colm.net>:
>>>>>>  Hi Konstantin,
>>>>>>
>>>>>>  code_t is the type used to encode virtual machine instructions.
>>>>>>  Just
>>>>>>  need 1 byte for this now.
>>>>>>
>>>>>>  half_t is used mostly for virtual machine instruction arguments. It
>>>>>>  started out as 2 bytes, but has grown. It might have been made
>>>>>>  larger
>>>>>>  than needed at the moment -- haven't looked at that in long time.
>>>>>>
>>>>>>  word_t is the primary type for virtual machine values pushed to the
>>>>>>  stack and operated on by the VM. The values are either pointers to
>>>>>>  things or integers. The VM assumes the same size here, so whatever
>>>>>>  type
>>>>>>  is used it needs to be the same size as a pointer or there will be
>>>>>>  some
>>>>>>  trouble.
>>>>>>
>>>>>>  Admittedly, portability needs to be improved.
>>>>>  Hi Adrian,
>>>>>  thank you for the clarification. I think this is enough for
>>>>>  experiments.
>>
>>  After studying the code, several attempts and experiments, I want to
>>  say that adding support for 64-bit platforms is not a trivial task.
>>
>>  Unfortunately I do not have the time and energy to solve this issue.
>>
>>>>>>  On 2018-04-30 11:34, Konstantin Podsvirov wrote:
>>>>>>>   30.04.2018, 18:06, "Adrian Thurston" <thurston at colm.net>:
>>>>>>>>   Hi, yes would like to have this fixed at some point. Sorry no
>>>>>>>>  ETA at
>>>>>>>>   the moment.
>>>>>>>
>>>>>>>   But can you briefly describe what does each type code_t, word_t,
>>>>>>>   half_t and what are the requirements in size and made to them.
>>>>>>>
>>>>>>>>   On 2018-04-29 12:53, Konstantin Podsvirov wrote:
>>>>>>>>>   Hello colm developers... :-)
>>>>>>>>>
>>>>>>>>>   Currently the word_t is hardcoded in src/tree.h file as
>>>>>>>>>  unsigned long
>>>>>>>>>   type, bun used to store pointers too.
>>>>>>>>>
>>>>>>>>>   On some platform (for example mingw64) size of the long (4
>>>>>>>>>  byte) less
>>>>>>>>>   then size of the void* (8 byte).
>>>>>>>>>
>>>>>>>>>   Previously i add check for this problem in CMake project file,
>>>>>>>>>  but
>>>>>>>>>   not provide any solution.
>>>>>>>>>
>>>>>>>>>   What you think about this?
>>>>>>>>>
>>>>>>>>>   --
>>>>>>>>>   Regards,
>>>>>>>>>   Konstantin Podsvirov
>>>>>>>>>
>>>>>>>>>   _______________________________________________ colm-users
>>>>>>>>>  mailing
>>>>>>>>>   list colm-users at colm.net
>>>>>>>>>   http://www.colm.net/cgi-bin/mailman/listinfo/colm-users
>>>>>>>>   ,
>>>>>>>>
>>>>>>>>   _______________________________________________
>>>>>>>>   colm-users mailing list
>>>>>>>>   colm-users at colm.net
>>>>>>>>   http://www.colm.net/cgi-bin/mailman/listinfo/colm-users
>>>>>>>
>>>>>>>   --
>>>>>>>   Regards,
>>>>>>>   Konstantin Podsvirov
>>>>>>>
>>>>>>>   _______________________________________________
>>>>>>>   colm-users mailing list
>>>>>>>   colm-users at colm.net
>>>>>>>   http://www.colm.net/cgi-bin/mailman/listinfo/colm-users
>>>>>>
>>>>>>  _______________________________________________
>>>>>>  colm-users mailing list
>>>>>>  colm-users at colm.net
>>>>>>  http://www.colm.net/cgi-bin/mailman/listinfo/colm-users
>>>>>
>>>>>  _______________________________________________
>>>>>
>>>>>  colm-users mailing list
>>>>>  colm-users at colm.net
>>>>>  http://www.colm.net/cgi-bin/mailman/listinfo/colm-users
>>>>
>>>>  Regards,
>>>>  Konstantin Podsvirov
>>
>>  _______________________________________________
>>  colm-users mailing list
>>  colm-users at colm.net
>>  http://www.colm.net/cgi-bin/mailman/listinfo/colm-users
>
> _______________________________________________
> colm-users mailing list
> colm-users at colm.net
> http://www.colm.net/cgi-bin/mailman/listinfo/colm-users

Regards,
Konstantin Podsvirov