[ale] RH 7.0 wth boraken compiler

[Benjamin Scherrey]

Here's the scoop...

	gcc 2.8.x was an attempt to redo a bunch of stuff and was fouled up
well enough (mainly in the C++ area which was the most important
changes) that it became useless to those wanting a standard's compliant
C++ compiler this century (kinda like RMS saying we should really wait
for HURD and Linux is just a temporary measure). This annoyed enough of
the talented gcc contributors enough that they split off at the gcc 2.75
branch and started the egcs. Lots of hurts feelings and no progress on
gcc and significant improvement with egcs later (finally became a pretty
good C++ compiler and still generated good enough C results to get Linus
to accept it to build the kernel) and RMS finally negotiated with the
egcs folks to take over complete development of gcc officially. egcs was
*NOT* merged to the old gcc, it was a code split. The 2.8.x branch was
abandoned, egcs was simply renamed gcc 2.95.x, and gcc 2.95.2 became the
first stable release of the new compiler system which is still far from
a complete re-write although it is an extreme improvement.

	RedHat, despite previously doing really smart things, screwed up by
releasing a patched version of the development tree, i.e. 2.96.x. The
reason is that binary compatibility is broken between 2.95 and 2.96
because of the new C++ ABI that was introduced but still unstable in
2.96. gcc 3.0 will be the first stable release of gcc with the new ABI
and a lot of other dramatic improvements in compliance. Anything between
2.95 and 3.0 will have to be completely rebuilt. An initial beta of 3.0
has just been released and I'm hoping that the stable release will be
declared before the end of the first quarter this year.

	The rest of the issues you mention about glibc are pretty much right
on. Hopefully RedHat will refrain from such sillyness in the future. I
still use RH 6.2 btw and look forward to the relase after gcc 3.0 is
available.

	later,

		Ben Scherrey
--
To unsubscribe: mail majordomo at ale.org with "unsubscribe ale" in message body.