2 Sep
2012
2 Sep
'12
4:03 p.m.
On 02/09/12 6:24 PM, Guy Sotomayor wrote:
On Sep 2, 2012, at 3:10 PM, Toby Thain wrote:
Not disputing any of this -- it's a natural cycle.
--Toby
On 02/09/12 5:31 PM, Dave McGuire wrote:
And it's a sorry looking code base. That's why clang started up?to replace the
mess that gcc had become. Free stuff is great, but the problem was no one was willing (or
able) to go back and clean it up. All any of the vendors (who were the ones adding
various CPU support into gcc) wanted to do was to make the smallest changes necessary to
get their CPU supported. The internals to gcc have needed replacing for years.
Unfortunately the internals are poorly documented and it's damn near impossible to
replace the pieces of it and still have the various front ends or code gens still work.
So folks started over from scratch. That's where clang came from (but admittedly to
solve some other issues with gcc).
...
Most highly technical organizations (or organizations with competent
technical staffs) had their own "standard OS load" for a given platform,
that consisted of a commercial vendor-supplied UNIX (SunOS, Ultrix, etc)
with all the bad stuff fixed. Fixing stuff usually meant replacing
"tar", "awk", etc with stuff that had actually progressed since the
1970s...and it was almost always GNU. Further, you had source code, so
you could fix any bugs you ran into without having to deal with talking
to incompetent vendor employees and MAYBE getting a bug fix in a few months.
Tell the kids these days - they won't believe yer.
Seriously: the great Reddit horde has no idea how bad things were before GNU shook it up.
It was hard to dream in 1985 that gcc would supplant just about every vendor compiler, and
when RMS said he hoped this would happen, many said he was nuts (we have Usenet archives
to prove it).
But it did eventually happen. :) TTFN - Guy