Re: uclibc support reference implementation, take 1
Hello Ron,
My 2 cents...
dpkg 1.14.4 from Sid. Source is unmodified except for the following:
ostable:
+ uclibc-linux linux-uclibc linux[^-]*-uclibc
triplettable:
+ uclibc-linux-<cpu> uclibc-<cpu>
For me, it looks sane. Probably we shall also add newlib-*, dietlibc-*, etc...
binutils 2.17cvs20070426-8 from Sid. This works unmodified to create
uclibc-arm packages using the new table entries above.
It probably fails with arches with 64 bit libraries, like powerpc.
Did someone/you test it?
gcc-4.1.2 from SLIND. This I've used partly because gcc-4.2 for arm
seems to have a few teething problems, and partly because they have
a nice mechanism for creating a packaged bootstrap compiler that isn't
supported in the mainline gcc source yet.
You meant Debian's upstream instead gcc mainstream, right?
It needed a few patches to
build, partly to fix broken debian/patches for arm in that particular
release, and partly to fix a SLIND assumption that DEB_TARGET_ARCH_OS
returns 'uclibc', which is not the case with the above table entries.
I'll send those patches to them separately.
From your point of view - as you have been playing with it - do you
think slind bootstrap could be ported to debian sanely and be kindly
accepted?
uclibc 0.9.29. This is a new upstream source, with the SLIND patch
and packaging wrapped around it.
And that's pretty much it for the first step. Toss them all together,
go brew a pot of coffee, and by the time your bloodstream is rushing
alkaloids and cream to the bits that most need them, you've got yourself
a(nother) uclibc toolchain.
:-)
What remains to be done:
- Ratify the new dpkg table entries here, and submit them for addition.
I guess the next step to that is for people to submit any objections
or suggested modifications they might have to the above. I know that
these values work now, but they still may not be the best to use ...
For me looks sane, but lets see what other power users might say.
Thanks for the work.
--
Héctor Orón
Reply to: