Re: flex beta 2.5.23 released

[Bruce Lilly]

W. L. Estes wrote:

> gcc does not provide the sys/types.h header file. that is provided by
> the system's C library.

I don't quite follow that; on a system with no bundled compiler (e.g.
recent Solaris), there are no headers unless supplied with a separate
compiler (such as gcc), no? [No, I'm not picking on Sun, the same
applies to MS Windows, which has no headers unless supplied with a
separate compiler.]

> Please list the platforms on which flex will not compile. Please send
> any configure output and such that will help us understand the nature
> of those systems.

For the record, do you want all of the output posted to the list or
off-list?  In the case of the FLEX_NEEDS... issue that means *two* sets
of configure output, two sets of config.* files, two sets of output
from make, etc. in addition to patches, description, and so on.


> > Agreed that would be a very bad idea.  But that's essentially what has
> > been done, since *some* mechanism to determine whether or not to compile
> > the flex-generated .c file with or w/o FLEX_NEEDS... will have to be used
> > on each system on which that file will be compiled -- it that mechanism
> > isn't exactly autoconf, it will be something equivalent.
>
>
> No. people compiling flex will have to manually add the -D
> option. That's simple and hardly as complex as autoconf.

It's per-system configuration nevertheless. For practical software
incorporating a lexical analyzer built by flex and distributed, that
essentially means something of complexity on the order of autoconf
since the makefile etc. that will be used to build the target software
which incorporates the flex-built .c file either will or won't need
to be compiled with a cc -D option, and that will vary from system to
system.

> I'd love it
> if it weren't necessary to include any mechanism to account for this,
> but it is.

It's not necessary w/ 2.5.4a and wasn't necessary with early betas; the
problem is strictly related to the change from char/short/int to intN_t.

> It's only necessary to pass the switch when building
> flex. once flex is built, flex does what it was told to do for your system.

Unfortunately that's not correct (at least for 2.5.23); each .c file
generated by fles 2.5.23 might or might not require a macro definition
(and that's what configuration is) to compile on each system on which the
flex-generated lexical analyzer will be built.