Re: Disabling lin-thread.c module

[Mark Kettenis <kettenis at wins dot uva dot nl>]

   Date: Thu, 13 Apr 2000 16:57:34 +1000
   From: Andrew Cagney <ac131313@cygnus.com>

   GDB 5.0 is all about getting new features and being able to boast about
   them.  GDB 5.0 is also all about kick starting the public CVS repository
   and getting real community development.  I suspect that people in
   mission critical situations will think twice about jumping on a new
   radical GDB 5.0 when:

	   o	they see an even/zero sub-version
		   (always a bad sign :-)

	   o	they know a 5.1 is coming soon

   With respect to thread_db, my understanding is that it is no less
   functional (1) then the old GDB-4.18 while at the same time, definitely
   being better architecturally designed (true?).  No matter how much
   testing effort is put in, it is always going to take a release before it
   becomes stable.  Enable it now, and 5.1 should be ok.  Enable it in 5.1
   and 5.2 (a year from now?) is likely stable :-/

Since GDB 4.18 didn't support threads on Linux GDB 5.0 will certainly
be `no less functional' :-).

Unfortunately people will compare 5.0 to HJ's "improved" 4.17 based
version.  I suspect that that version still does handle threads
somewhat better than the current code in 5.0.  But I guess our only
chance to improve things is getting the code out, and hope that people
will dump HJ's version when they discover all the nifty new features
in GDB 5.0, and see that we're actually fixing the bugs thet reported :-).

Anyhow, the strongest argument for shipping the libthread_db code is
that this will put an end to the practice of fixing GDB over and over
again when the internals of the LinuxThreads library are changed.

   The decision is between the thread and Linux maintainers.  The thread
   maintainer said ya, so linux maintainers?

That's a `ja' from me to :-).

Mark