Re: Patch not included in e2fsprogs 1.21.

2001-06-19 Thread Adam Di Carlo
Yann Dirson [EMAIL PROTECTED] writes: As for the packaging of those libs and of the binaries that were reduced, I'm not sure what would be best. IIRC most stuff included in boot-floppies is taken right from real packages installed on the machine where they are built. The e2fsprogs-bf I've

Re: Patch not included in e2fsprogs 1.21.

2001-06-19 Thread Yann Dirson
On Tue, Jun 19, 2001 at 01:41:53PM -0400, Adam Di Carlo wrote: I don't really see the problem with having an e2fsprogs-bf package, just for root.bin, which no one should ever normally install... Well, my fear was that people would install e2fsprogs-bf on their system, but now that I think of

Re: Patch not included in e2fsprogs 1.21.

2001-06-17 Thread Philip Blundell
Isn't PIC used for a reason in shared libs? :) Sure it is, but that reason doesn't really apply here. The reason we compile shared libraries as PIC is so that many applications can map the same object at different addresses and have only one copy of the code in memory. PIC code is slightly

Re: Patch not included in e2fsprogs 1.21.

2001-06-17 Thread Philip Blundell
Uh, wrong. We compile _all_ shared libs using -fPIC (IOW, every .so). That has nothing to do with mklibs.sh, it has to do with being a shared lib. If you try to link a .so with non-PIC objects, some architectures will barf all of the resulting library. The glibc testsuite checks that it works to

Re: Patch not included in e2fsprogs 1.21.

2001-06-17 Thread Philip Blundell
I understand this is policy -- specifically policy.text.gz, 4.2. This policy makes perfect sense for a.out shared libraries (where were all the rage when this policy was written). Does this policy still make sense? Actually, this is a bit backwards. Linux a.out shared libraries never

Re: Patch not included in e2fsprogs 1.21.

2001-06-17 Thread Theodore Tso
On Sun, Jun 17, 2001 at 12:27:57PM +0100, Philip Blundell wrote: Isn't PIC used for a reason in shared libs? :) Sure it is, but that reason doesn't really apply here. The reason we compile shared libraries as PIC is so that many applications can map the same object at different addresses

Re: Patch not included in e2fsprogs 1.21.

2001-06-17 Thread David Whedon
The other issue is that the lite libext2fs.so.2.4 should never ever be installed on a normal system, since it's missing a lot of necessary object files. In fact, when I was chatting with Larry McVoy, he warned me that I might be unloosing a time-bomb that might come back to haunt me. My

Re: Patch not included in e2fsprogs 1.21.

2001-06-16 Thread Yann Dirson
[cc-ing boot-floppies team to let them know] On Fri, Jun 15, 2001 at 07:44:43PM -0400, [EMAIL PROTECTED] wrote: I've rejected this patch, for two reasons. (1) It's not needed any more since you can now configure e2fsprogs to generate a lite version of libext2fs.so.2.4. OK I'll replace

Re: Patch not included in e2fsprogs 1.21.

2001-06-16 Thread Philip Blundell
(2) The patch is busted (not correct). The object files being placed into the libext2fs_pic.a file are the non-shareable .o files, not the PIC .o files. You need to cd into the elfshared subdirectory to get at the PIC files, and the Makefile fragment doesn't do that. Woops... I

Re: Patch not included in e2fsprogs 1.21.

2001-06-16 Thread Ben Collins
On Sat, Jun 16, 2001 at 02:28:08PM +0100, Philip Blundell wrote: (2) The patch is busted (not correct). The object files being placed into the libext2fs_pic.a file are the non-shareable .o files, not the PIC .o files. You need to cd into the elfshared subdirectory to get at

Re: Patch not included in e2fsprogs 1.21.

2001-06-16 Thread Philip Blundell
Wouldn't building the .so with non-PIC break quite a few things? Shouldn't do, as far as I can see. Anything in particular you are thinking of? p. PGP signature

Re: Patch not included in e2fsprogs 1.21.

2001-06-16 Thread Yann Dirson
On Sat, Jun 16, 2001 at 02:28:08PM +0100, Philip Blundell wrote: Actually, for boot-floppies' purposes it would be slightly better to build the shared library with non-PIC objects. You may not have done it intentionally but it sounds like you got the right result. :-) Hm... let me know

Re: Patch not included in e2fsprogs 1.21.

2001-06-16 Thread Erik Andersen
On Sat Jun 16, 2001 at 10:52:25AM -0400, Ben Collins wrote: On Sat, Jun 16, 2001 at 02:55:30PM +0100, Philip Blundell wrote: Wouldn't building the .so with non-PIC break quite a few things? Shouldn't do, as far as I can see. Anything in particular you are thinking of? Isn't PIC used

Re: Patch not included in e2fsprogs 1.21.

2001-06-16 Thread Ben Collins
On Sat, Jun 16, 2001 at 12:53:08PM -0600, Erik Andersen wrote: On Sat Jun 16, 2001 at 10:52:25AM -0400, Ben Collins wrote: On Sat, Jun 16, 2001 at 02:55:30PM +0100, Philip Blundell wrote: Wouldn't building the .so with non-PIC break quite a few things? Shouldn't do, as far as I can

Re: Patch not included in e2fsprogs 1.21.

2001-06-16 Thread Erik Andersen
On Sat Jun 16, 2001 at 03:04:28PM -0400, Ben Collins wrote: Uh, wrong. We compile _all_ shared libs using -fPIC (IOW, every .so). That has nothing to do with mklibs.sh, it has to do with being a shared lib. If you try to link a .so with non-PIC objects, some architectures will barf all of the

Re: Patch not included in e2fsprogs 1.21.

2001-06-16 Thread Ben Collins
On Sat, Jun 16, 2001 at 01:22:50PM -0600, Erik Andersen wrote: On Sat Jun 16, 2001 at 03:04:28PM -0400, Ben Collins wrote: Uh, wrong. We compile _all_ shared libs using -fPIC (IOW, every .so). That has nothing to do with mklibs.sh, it has to do with being a shared lib. If you try to link a

Re: Patch not included in e2fsprogs 1.21.

2001-06-16 Thread Erik Andersen
On Sat Jun 16, 2001 at 03:26:59PM -0400, Ben Collins wrote: We have _always_ required (by policy) that objects for shared libs be compiled with -fPIC. Where have you been? That's the whole definition of a shared library is that it is relocatable. One of the many beautiful things about ELF

Re: Patch not included in e2fsprogs 1.21.

2001-06-16 Thread Ben Collins
On Sat, Jun 16, 2001 at 02:49:32PM -0600, Erik Andersen wrote: PIC code has a major performance impact since one entire register is devoted exclusively to holding the base address. The performance impact is exacerbated on lame arches such as x86 which have far too few registers. Only