[Scilab-Dev] linux 64 bits support

[Sylvestre Ledru]

Nice work you have done Simon !

>  - Reducing stack (e.g. stacksize("min")) doesn't work (that's why
> nonreg test of bug #2867 fails). Since it's rarely used (seems to
> never be done automatically by scilab) and requires big changes (far
> more changes than this patch) in functions I don't understand, it's
> likely no worth fixing it. Unless I find a simple way...
Not very important bug indeed. Don't need to spend time on this.

>  - Funny part: "make clean" fails (segmentation fault). I'll try to see
> this tomorrow.
hmhm, does it still happen ? 
"make clean" is calling Scilab to clean the documentation, maybe it is
where your bug occurs.

>  - Another funny thing (no link with 64 bits): when running "make
> check" on a distant machine with X forwarding, some tests fails but in
> the report "make check" says that two extra tests passed (103 passed in
> local, 105 passed with X forwarding and some errors). More errors =
> less errors... do less with more ? ;)
hmhm, funny :)

> Results of "make check", by module:
>  - Same results on 32 bits and 64 bits versions
> arnoldi, boolean, completion, elementary_functions, genetic_algorithms,
> integer, output_stream, overloading, parameters, polynomials,
> signal_processing, simulated_annealing, sparse, special_functions,
> spreadsheet, statistics, symbolic, m2c, cascd, data_structures,
> differential_equations, dynamic_link, io, linear_algebra, optimization,
> randlib (-- and all modules  which doesn't have any tests ;))
> 
>  - Different results:
> compatibility_functions: bug_1227 fails on 64 bits and not on 32 bits
> version. Both hangs on bug_577. Don't know why.
> double: 25% passed, 75% in 32 bits version. Failed tests are trying to
> reduce stack size, so that's not a big issue
> fileio: 94% passed, 100% in  32 bits version. Don't know why.
> functions: bug_625 and bytecode pass fine on 32 bits but not on 64 bits
> matio: 64 bits version freeze, 32 bits version fails. Don't know why.
> sound: sound doesn't passes on 64 bits. Don't know why.
> string: strcmp and *pcre76 doesn't passes. Don't know why.
> time: getdate and bug_793 doesn't passes. Don't know why, but Allan did
> special things on Windows 64 bits on this module, I'll also see this
> tomorrow.
OK, it is just great seeing a 64 bits version coming back from the dead.

Sylvestre