[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: real LSB compliance

On Sat, Jun 30, 2001 at 10:51:28PM -0700, Daniel Jacobowitz wrote:
> On Sun, Jul 01, 2001 at 03:28:31PM +1000, Christopher Yeoh wrote:
> > 
> > LSB compliant apps will need to link against /lib/ld-lsb.so.1. This
> > doesn't mean that Debian distributed packages have to do this but the
> > loader will have to be available.
> [trying not to start an argument, I'm sure this has been discussed at
> length, but...]
> What on earth is this supposed to accomplish?  As far as I know, every
> reasonably current Linux distribution for any particular architecture
> agrees on the name of the dynamic linker.  Yes, it's different per
> platform - because it was necessary to bump the version after
> particular mistakes or ABI improvements on some particular
> architecture.  But this isn't a rabid compatibility issue, and as
> things stand no one ever needs to twiddle that, and the system compiler
> will always be expected (perfectly reasonably) to get it right.  Why
> introduce a gratuitous complication in the compilation of any
> LSB-compliant program, and a gratuitous dependency on a currently
> nonstandard linker when the existing one seems entirely adequate?

As a random guess, probably so that you can fully support LSB
apps without limiting your native apps to whatever strangeness
the LSB might declare.  Or, for the typical vendor, so that
you can fully support LSB without fixing your brokenness.
Assuming that the native ld-linux.so.2 handles LSB apps
correctly, a link from ld-lsb.so.1 to ld-linux would probably

I'm further guessing that this is a result of two vendors
fighting about who's libc soname is correct.  It's completely
stupid, like some of the details of most standards.


Reply to: