[petsc-dev] [petsc-maint #50087] Re: why does PETSc BuildSystem hate me?

[Barry Smith]

  Aron,

   Your complaints about PETSc's ./configure are legitimate, there is no documentation and the --help spews a bunch of incoherent, badly formatted and inconsistent information. In addition there is no documentation about parts of autoconf it supports versus what parts of autoconf it does not support.

   I have gone through all the help strings to use more consistent notation and improve the formatting; I think --help is much better now in petsc-dev. 

   It would be possible to generate a .html version of the help automatically, but for it to be useful we should add more information into the .py that describe the options more fully. Let's think about this.


   Barry


On Jul 26, 2010, at 1:00 PM, Aron Ahmadia wrote:

> Barry and Satish,
> 
> While I have your eyes, we should probably discuss what the ideal way for me
> to tell PETSc how my system is set up.
> 
> As you can probably tell, I am doing a lot of the user environment
> configuration for our BlueGene using modules and some Python scripting to
> back them up, so softenv+some more magic on the administrator side to
> streamline the process.  This isn't my full time job, but I'd like to do it
> right.
> 
> I like the idea of telling PETSc how to use each of the packages, but right
> now the usage is not well-defined.  There's a huge dump of available options
> if you type configure -h but no good documentation on the .py files that I
> know of.  In fact, it might be useful to be able to generate an HTML page
> that lists all the options and what they do.
> 
> I think I have gotten things working, both by careful .py file definitions
> and also by jamming all of my link flags and libraries into LIBS.  Perhaps
> we should break them in to LDFLAGS and LIBS?  I am okay with explicitly
> giving PETSc my intended environment variables, this is safer for everyone.
> 
> 
> If there's a way I can help on this, let me know.  I need to remember to
> only bang my head against PETSc problems when you guys are awake :)
> 
> A
> 
> On Mon, Jul 26, 2010 at 8:45 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:
> 
>> 
>> Satish,
>> 
>>  It is missing Aron's favorite. Could you please add it LDFLAGS
>> 
>>  I was expecting it up in config/configure.py before anything else? Won't
>> that be a better place for it?
>> 
>> 
>>  Barry
>> 
>> On Jul 26, 2010, at 12:13 PM, Satish Balay wrote:
>> 
>>> On Mon, 26 Jul 2010, Barry Smith wrote:
>>> 
>>>> 
>>>> On Jul 26, 2010, at 12:04 PM, Satish Balay wrote:
>>>> 
>>>>> Barry,
>>>>> 
>>>>> I've added code in petsc-dev that warns users of the env variables
>>>>> [does not ask them to unset]. It then ignores these values - and never
>>>>> uses them in configure process.
>>>> 
>>>> Where is the check, I cannot find it anywhere?
>>> 
>>> resetEnvCompilers() in setCompilers.py.
>>> 
>>> Perhaps we need to add more variables in that list..
>>> 
>>> Satish
>>> 
>> 
>> 
> 
> Barry and Satish,
> 
> While I have your eyes, we should probably discuss what the ideal way for me to tell PETSc how my system is set up.
> 
> As you can probably tell, I am doing a lot of the user environment configuration for our BlueGene using modules and some Python scripting to back them up, so softenv+some more magic on the administrator side to streamline the process.  This isn't my full time job, but I'd like to do it right.
> 
> I like the idea of telling PETSc how to use each of the packages, but right now the usage is not well-defined.  There's a huge dump of available options if you type configure -h but no good documentation on the .py files that I know of.  In fact, it might be useful to be able to generate an HTML page that lists all the options and what they do.
> 
> I think I have gotten things working, both by careful .py file definitions and also by jamming all of my link flags and libraries into LIBS.  Perhaps we should break them in to LDFLAGS and LIBS?  I am okay with explicitly giving PETSc my intended environment variables, this is safer for everyone.  
> 
> If there's a way I can help on this, let me know.  I need to remember to only bang my head against PETSc problems when you guys are awake :)
> 
> A
> 
> On Mon, Jul 26, 2010 at 8:45 PM, Barry Smith <bsmith at mcs.anl.gov> wrote:
> 
>  Satish,
> 
>   It is missing Aron's favorite. Could you please add it LDFLAGS
> 
>   I was expecting it up in config/configure.py before anything else? Won't that be a better place for it?
> 
> 
>   Barry
> 
> On Jul 26, 2010, at 12:13 PM, Satish Balay wrote:
> 
> > On Mon, 26 Jul 2010, Barry Smith wrote:
> >
> >>
> >> On Jul 26, 2010, at 12:04 PM, Satish Balay wrote:
> >>
> >>> Barry,
> >>>
> >>> I've added code in petsc-dev that warns users of the env variables
> >>> [does not ask them to unset]. It then ignores these values - and never
> >>> uses them in configure process.
> >>
> >>  Where is the check, I cannot find it anywhere?
> >
> > resetEnvCompilers() in setCompilers.py.
> >
> > Perhaps we need to add more variables in that list..
> >
> > Satish
> >
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20100726/81d2b2dc/attachment.html>