Re: [Gcl-devel] Re: describe bug, was: Re: [Maxima] First step toward a 5.9.2 release

[Camm Maguire]

Greetings!

Well, now that we have our own version in place, we certainly
shouldn't make life difficult for any of the ports.  My original
thinking was to offload another bit of code, and transparently capture
improvements as they are made externally.  

Tre does look very interesting, expecially its memory allocation
strategy.  One caviat I had about externalizing regexp was that allocs
would then take up slow contiguous blocks, while with the inlined
version I can specify relocatable regexps and ensure that regexec does
not trigger a gc.  

In general, if you feel this would cause any headaches, lets just
leave well enough alone -- I doubt anyone cares much about regexp
efficiency, yes?

Take care,


"Mike Thomas" <address@hidden> writes:

> Hi Camm.
> 
> Thanks as always for your tremendous efforts.  I'm almost completely
> ignorant of the pros and cons of regex libraries but these are my thoughts:
> 
> | 1) libc POSIX regex -> eliminate old o/regexp.c
> 
> Note that although GNU regex and pcre etc are available for MinGW32 they are
> separate items that intending Windows developers have to chase down to make
> a complete set of GCL development tools.
> 
> I haven't tracked down the licences of these items yet but also bear in mind
> that another strict GPL licence is another nail in the coffin for an LGPL
> GCL.
> 
> Also see http://laurikari.net/tre/index.html which looked interesting.
> 
> Cheers
> 
> Mike Thomas.
> 
> 
> 
> 
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah