Re: [boost] Boost.Build on IBM, SGI & HP

From: Toon Knapen (toon.knapen_at_[hidden])
Date: 2001-10-10 13:13:27

David Abrahams wrote:

>>On SGI, jam compiled fined (after setting the executable flag on
>>'yyacc'). Some time ago, you agreed with my proposition to provide
>>jam-executables to the boost users for the most popular platforms. Do I
>>already check in the SGI version and where ?
>>
> 
> I'm not sure yet. What do you think?

Well you know my point of view. Boost Users should be able to build 
boost at their site with a minimum of hassle. If the installation is not 
  configure ; make ; make install, it should at least be as simple. 
Including the jam-executable in the boost download package helps to make 
local installation simple.
Maybe we should ask Boost-Users what they prefer ?

>>On HP (HP-UX 10.20) I did not succeed in building 'jam0' since 'alloca'
>>is an undefined symbol. I have searched through all system libraries but
>>could not find a definition of alloca. The man page directed me to an
>>old library which neither defined 'alloca'. I found some information on
>>the web though confirming that 'alloca' is indeed missing on (some / all
>>?) HP-UX machines together with an open-source implementation of the
>>routine.
>>(ref :
>>
>>
> http://www.informatik.tu-muenchen.de/rechenbetrieb/infohalle/progber/Portier
> ung/alloca.html)
> 
> OK, the use of alloca is part of code in jamgram.c, which is generated by
> gnu bison, the only version of yacc that I have on my system. If you look at
> the previously-checked-in version of jamgram.c, which was generated by some
> other yacc, you'll see that there's no use of alloca. If there is a yacc on
> your HP machine, it could be used to generate a new jamgram.c, which, I
> imagine, would not use alloca.

Funny thing is that I've built everything locally, so I used the yacc 
provided by HP.

I'll keep you up to date about trying out the workarounds next time I've 
access to the IBM and HP (only few days a week)

toon