On Saturday 01 November 2008 07:26:04 Christian MICHON wrote:
> Jivin David McCullough lays it down...
>
> > I don't recall any special magic
>
> Some magic could be invoked to save time at least :-)
>
> More investigations on D*B, gcc-3.4.6 and uclibc-0.9.29, trying to
> compile stock uClibc++-0.2.2.
>
> The only use of libsupc++.a is apparently to extract some objects from it
> only.

Er, yes, I said:

> All I really _need_ is a half-dozen .o files bundled into that .a file,

> so we could theoretically replace this extraction and dependency to
> libsupc++ by just compiling in place these objects (extracting the
> source code from the exact g++ version you're using: in my case,
> gcc-3.4.6).

Hence:

> *shrug*  I could beat compilation out of it by hand, but I'd like something
> that has at least a minimal chance of working in future gcc versions...

I'm currently fiddling with gcc 4.1.2, and would like to move on to 4.2 or 4.3 
once I have time to attack the arm soft float static linking issue.  Which is 
after the 0.9.30 release, and after I get my server doing decent nightly 
snapshots automatically.  (It's doing nightly snapshots, but I'm not happy 
with 'em yet.)

Speaking of which, how's e release going?  (Are the stars right?)

> In this way, Rob could bypass the long and painful compilation of
> g++/libstdc++ by just compiling these version-dependent files from
> g++. This would mean dedicated patching of Makefile (the variation is
> the version of gcc).

Which is evil and nasty.  (Possibly less so than building the whole of 
libstdc++, but still.)

The fundamental problem is that the gcc build is very non-orthogonal.

If I can cd into the libsupc++ directory and get it to build just that, I'll 
probably be happy with that.  (It's a separate directory, it has its own 
makefile.  Logically this would imply _some_ kind of separation from the rest 
of the build.  But logic and the gcc build process seem to have very little 
in common so far, other than perhaps a distant ancestral relationship...)

Writing my own makefile that feeds in the appropriate hand-generated .config 
stuff sounds like a maintenance nightmare.

> thoughts, comments ? at least, I'll try that myself on a native system
> tonight... :)

Let me know how it goes...

Rob
_______________________________________________
uClibc mailing list
[email protected]
http://busybox.net/cgi-bin/mailman/listinfo/uclibc

Reply via email to