Re: Reentrant flex

[Bojan Smojver <bojan rexursive com>]

Quoting "Michael A. Peters" <mpeters mac com>:

> Maybe one of the FHS gurus has a better idea, but what I would suggest
> is
>
> a) trying to convince core maintainer to switch to it (might be
> hard/impossible)

That will be very hard, as this package isn't well maintained at all 
(last released March 31, 2003). There are some patches for it that 
Debian folk keep and I plonked those into the package I built. But a 
bigger problem here is that this version of flex is (was) supposed to 
become flex 2.6, therefore being the new official flex, as it has all 
the functionality of regular flex plus reentrant C bits. So, I'm 
guessing that upstream folk want to keep it they way it is. 
Unfortunately, nobody knows any more if 2.6 is going to happen and when.

> b) patching flex-reentrant to use a non conflicting name for header
> file, static lib, info page, etc.

I can do that and it shouldn't be that hard (there is only a bunch of 
files in the build I need to keep, the rest exists in regular flex, so 
it can be thrown away). Also, that will make sure that nobody uses that 
package instead of regular flex by accident. Given that scanners are 
vital to proper functioning of many packages, accidentaly replacing the 
old flex with this one would be a disaster and I certainly don't want 
to do that.

> Then in your mod-spin package, configure could first check for the
> modified flex-reentrant and use it - and if that's not there, check to
> see if the system flex is flex-reentrant.

I have options that are passed to configure that do that already. They 
would have to be tweaked somewhat, but I don't mind doing that. So, 
this is a non-issue. BTW, the whole thing is only important at build 
time anyway, so it shouldn't affect runtime at all.

Thanks for your suggestions. I'll report back when I have something 
better (i.e. code :-)

--
Bojan