[oi-dev] Libtool 2.4.2 and several questions (nvidia, build process)

[Alexander Pyhalov]

Hello.
I've just updated libtool to 2.4.2 version. It is a version from 
upstream with libltdl 1.5.22 compatibility library, so it should be 
safe.

I've tested this version of libtool (and libltdl)
a) with libtool test suite
- The only fails are related to the absence of libgfortran.so in 
standard library paths
This issue is easy to solve - we need to provide 
system/library/gfortran-4-runtime package - I think it's a matter of 
adding one *.p5m file to gcc-47 package
b) I've rebuilt illumos-gate with new libtool and libltdl installed ( I 
already know, this update shouldn't influence illumos-gate build process 
at all, but just to be sure)
c) I've rebuilt oi-userland with new libtool. This highlighted several 
problems with our packages and, possibly, build process.

1) Not all packages could be built in zone with shared ip stack (fixed 
this)
2) There was something strange with pixz component  - it was built on 
build server but I don't know how - while building locally it failed (so 
I updated it).
I think that this shows the necessity of periodic clean build of all 
packages (new and old) on build server.
Some packages has non-trivial dependencies - e.g., now automake depends 
on perl-510, however once rebuilt it will depend on perl-516. It was 
affected by other packages in a way that's hard to estimate 
automatically.

Andrzej, is it possible to make a clean build, for example, once a week, 
to catch some implicit interdependencies ?

3) gmake clean fails on components/nvidia.
Does someone have any idea why it needs this ( 
https://github.com/OpenIndiana/oi-userland/blob/oi/hipster/components/nvidia/Makefile 
):

clean::
	[ -d $(BUILD_DIR) ] && rm -rf $(BUILD_DIR)

Shouldn't this be transformed just to

clean::
	rm -rf $(BUILD_DIR)

?
-- 
System Administrator of Southern Federal University Computer Center