Re: Some bugs?

[Bill Allombert <Bill.Allombert@math.u-bordeaux1.fr>]

On Sat, Jul 04, 2009 at 08:31:40PM +0100, Jason Moxham wrote:
>
>> On Sat, Jul 04, 2009 at 12:24:00PM +0100, Jason Moxham wrote:
>>> On Thursday 02 July 2009 23:11:31 Jason Moxham wrote:
>>> I manage to get rid of a few more bugs, and I'm left with these
>>>
>>> polred,rnf,rnfkummer all have this error
>>>
>>> x^7 + Mod(7*y - 7, y^2 - y - 1)*x^6 + Mod(-21*y + 28, y^2 - y - 
>>> 1)*x^5 - 35*
>>> x^4 + Mod(35*y - 49, y^2 - y - 1)*x^3 + Mod(-7*y + 84, y^2 - y - 
>>> 1)*x^2 + Mo
>>> d(-14*y + 21, y^2 - y - 1)*x + Mod(-y - 43, y^2 - y - 1)
>>>   ***   at top-level: rnfpolredabs(nfinit(
>>>   ***                 ^--------------------
>>>   *** rnfpolredabs: could not open requested file ./MPQS.gpa/FREL.
>>
>> There might a problem with the way MPQS creates temporary files.
>> Try factor(2^128+1)
>> The code to create temporary files is in src/language/es.c.
>> Maybe we use a wrong directory or a wrong filename.
>>
>
> Yes it is , mpqs assumes the function below creates the directory as well 
> as testing existence
>
> line 4175 in language/es.c
> static int
> pari_dir_exists(const char *s) { return 0; }
>
> change to
>
> static int
> pari_dir_exists(const char *s) { return mkdir(s); }
>
> So this is the same as the UNIX one but without the mode
> So I suppose we will need this for MSVC and MinGW

Well, the code is fundamentally flawed. This need to be improved.
There is no excuse for a function 'pari_dir_exists' to create a directory.
We should improve that part.

> this fixes polred,rnf,rnfkummer  tests for Win32 MSVC

Good!

Bill.