[petsc-users] configure error with netcdf

[Barry Smith]

On Sep 9, 2012, at 12:02 PM, Satish Balay <balay at mcs.anl.gov> wrote:

> On Sun, 9 Sep 2012, Barry Smith wrote:
> 
>> 
>> On Sep 9, 2012, at 7:55 AM, Matthew Knepley <knepley at gmail.com> wrote:
>> 
>>> On Sun, Sep 9, 2012 at 12:31 AM, Satish Balay <balay at mcs.anl.gov> wrote:
>>>> ncgentab.c:(.text+0x1e2): undefined reference to `yyunput'
>> 
>>   Satish,
>> 
>>  Weird undefined symbols like yyunput are always a result of flex issues. 
>> 
>>    Could you please make the same changes to netcdf.py that you made to the other package that used flex (which resolved the problem there).
> 
> In this case netcdf has its own configure - and does its own search for flex etc.
> 
>>>> 
> checking for flex... flex
> checking lex output file root... lex.yy
> checking lex library... -lfl
> <<<
> 
> So the tools appear to be found. And I don't see them being used
> further down [in configure.log].

   Agreed. It seems some of the processing is lost; perhaps netcdf build tools do not properly display everything they are doing

   Note the use of .l files and the symbol yyunput which is very specific to flex. 

>  So its not clear if 'missing flex'
> is the cause of this error.

   Yes, it may not be a "missing" flex that is causing the problem but the problem is related flex, bison and all that crap. Unfortunately if we cannot reproduce on a machine we have access to there is no way we can debug it.

    Barry

> 
> 
> And after my patch [with netcdf upgrade] - I have a successful build
> of netcdf on rhel5 - even without flex/bison.
> 
> However flex is needed for ptscotch anyway [so one would get an error
> from ptscotch if flex is missing]
> 
> Satish