This is the mail archive of the mailing list for the glibc project.

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

Re: __sysconf and tcltk 8.3

   From: Ulrich Drepper <>
   Date: 05 Nov 2000 13:45:23 -0800

   Andreas Jaeger <> writes:

   > You exported it with GLIBC_2.2.  What other version are you going to
   > use?  I don't see a reason to change this.  Can you explain the
   > reasons a bit more, please?

   The versioning scheme will not catch the case when a program linked
   with a new version is run on a system with 2.1.94.

Not immediately, although eventually, when the symbol is actually used
in the program it will print a pretty obvious error message.

Too bad for those people running a system with 2.1.94.  The GLIBC_2.2
version symbol is for symbols introduced in glibc 2.2.  Breaking that
correspondence is IMHO a very bad thing to do since it'll cause quite
a bit of confusion.

I think people running one of the pre-releases should upgrade as soon
as glibc 2.2 is officially released.  The fact that Red Hat included
2.1.94 in its Red Hat Linux 7.0 shouldn't make any difference.  They
should advise their costumers to upgrade to glibc 2.2 as well, and
provide the necessary updates to them.  They made a choice, they
should suffer the consequences, and I hereby volunteer to answer all
the mails from people complaining about a missing __sysconf@GLIBC_2.2
(telling them to upgrade to the real glibc 2.2).

Of course, if you want to avoid a conflict with your employer, you can
always revert the __sysconf()/CLK_TCK changes, release glibc 2.2, add
the stuff back (exporting __sysconf() with GLIBC_2.2.1) and release
glibc 2.2.1.  But we really need to keep the consequent 1:1
correspondence between the versioning symbols and officially released
glibc versions.


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