flex-2.5.31

[tushar@…]

Thanks to torsten for reporting a new version.

​http://sourceforge.net/project/showfiles.php?group_id=72099&release_id=135412

This bug should be added as a dependency to Bug 30.

[highos@…]

So has anyone been testing this new version? I had heard of it a while ago [from #lfs, just been too lazy to add ;p] and from everything i've heard, it does *not* play nice with default build instructions. I suggest anyone that's playing with it, to please pipe up on lfs-dev and report back your findings.

[tushar@…]

Version Increment (2.5.31).

BTW, I had no problems with 2.5.27.

[markh@…]

Just to add my thoughts to this, we should probably either update to 2.5.31 or at least patch 2.5.4a as the latter generates incorrect C++ code which causes things to bomb out during compilation when using gcc-3.*. I mention this as I'm compiling Lilypond and had this problem - solution was to grab the patch for flex from debian stable,recompile flex-2.5.4a and add using std::istream; using std::ostream; to /usr/include/FlexLexer.h. I didn't try it with 2.5.31 but I assume it'll have been fixed in this release.

[markh@…]

Apologies - my previous comment was incomplete. I also had to add:

using std::cin; using std::cout; using std::cerr;

to /usr/include/FlexLexer.h

It looks like it will definately be easier just to update Flex :-)

[Matthew Burgess]

But not to 2.5.31 Mark, there's problems with it that apparently the maintainer is aware of - it causes HJL's binutils to not compile but more importantly modutils doesn't compile either - see ​http://archive.linuxfromscratch.org/mail- archives/lfs-support/2003/04/0182.html.

2.5.27 seems alright, although by the time the package updates are done and the next book released then 2.5.31's problems may have been resolved.

Matt.

[markh@…]

In that case should we just adapt the debian patch for 2.5.4a to paper over the problems until there's a new flex release?

[Matthew Burgess]

Actually Mark, I'm not sure whether this is a similar thing to when bison got all standards compliant on us and broke other (non-standards compliant) packages. It may be that flex-2.5.31 is actually correct and it's just that HJL's binutils and modutils have broken code in them that the newer flex can't/won't handle. I've not tried flex-2.5.31 out myself so can't comment on the nature of the failures for definite, sorry.

Matt.

[tushar@…]

2.5.27 works well.

[Matthew Burgess]

See ​http://tinyurl.com/9m2o for a description of the problem with flex-2.5.31 and ​http://tinyurl.com/9m2i for the associated patch (a one line fix!). This should appease those wanting to live on the bleeding edge (including myself), although I've not tried the patch yet.

[gerard@…]

Don't use 2.5.31, too many problems. 2.5.27 works fine reportedly.

[gerard@…]

On April 28, 2003 04:16 pm, Zack Winkles wrote:

The audit trail for this bug says that flex 2.5.31 works fine with the patch. If it works, we might as well use it, right?

If the patch proves to work to build X and modutils, then sure we might as well use it then. I've heard that it might not work to build those packages, but I have no first hand experience.

[highos@…]

Version increment (2.5.31)

[gerard@…]

I kind of decreased that to .27 for a good reason. If you have additional infomration that one of those .31 patches work now, then please provide this info in forms of the actual path or links explaining those reported problems were unfounded, or what.

[gerard@…]

We'll deal with this later when a better patch has been found.

[jeremy@…]

We need to try to figure this one out before 5.0 release, so I'm going to set to P1. My experience has been that .27 works great, but .31 has flaws.

[Matthew Burgess]

I posted a patch to patches@… yesterday. It fixes the issue with modutils although I'm not sure about the other known breakage (XFree86). If someone wants to do a build with the patch applied I'd be greatful. Failing that then some error logs would be useful.

[tushar@…]

I did a build recently with flex-2.5.31 without any major hiccups.

[jeremy@…]

Because 2.5.31 is still listed as beta, we've decided to stay with 2.5.4a, unless a good reason can be found to upgrade.

[tushar@…]

FYI, the patch from Redhat RPM is available on the patches website.