This is the mail archive of the libc-hacker@sources.redhat.com mailing list for the glibc project.

Note that libc-hacker is a closed list. You may look at the archives of this list, but subscription and posting are not open.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: My proposal for the libgcc runtime ABI (ia64 gcc/glibc is broken.)


On Wed, Jul 12, 2000 at 12:48:12PM -0400, Phil Edwards wrote:
> 
> > Jeffrey A Law <law@cygnus.com> writes:
> > > Not necessarily.  One could certainly argue that if it's on the root
> > > filesystem that it ought to be linked statically.
> >
> > Jeff, don't be ridiculous.  First, you should look at the reality.
> 
> Well... he is.  I speak as a system administrator now rather than a
> developer; Jeff is quite correct in saying that this is an arguable point.
> It is by no means ridiculous.
> 
> I've had Linux boxes crash hard, and IRIX boxes crash hard, and Solaris,
> and Alphas; and Linux has been the only one ever that was a nightmare to
> repair due to all the dynamic linking tangles.  Having root-partition
> binaries be statically linked is /excellent/ preparation for problems.
> Or even just for doing work in a /real/ single-user mode.

I have been working on X11, kernel and libc for years. My Linux
machines have crashed more than once. I have never found dynamic
binaries under / is a big problem. Yes, it is different. You have to
have some tools around when it happens. I have recuse floppy and CD.
They are very useful to me.

> 
> Oh, we'll all be bitching at the GCC maintainers anyway even if they get it
> all correct.  :-)  But I'd rather have a delayed, well-though-out solution
> instead of a quick "this works perfectly for Linux, the rest of y'all are
> screwed for another 18 months".

All we suggest is

1. Define a libgcc runtim ABI and stick to it.
2. Leave the libgcc.so issue to the target system people.

Trust us. #2 is a very hard problem.

> 
> GCC is finally getting some /serious/ recognition on non-Linux platforms.
> Even non-programmers have heard of it and recognize that it is a quality
> compiler, but many of them still incorrectly think of it as "the Linux
> compiler" because that's where the attention goes.  Please, folks, let's
> not blow this now.

That is Ulrich's point and mine. To include libgcc.so in gcc, gcc has to
spend HUGE amount of sources on Linux alone just to get it right on Linux.
By leaving it to each system, gcc won't take all the blames when bad things
happen :-).



H.J.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]