Re: Assume solib.h

[Daniel Jacobowitz <drow at false dot org>]

On Fri, Nov 12, 2004 at 10:51:07AM -0500, Andrew Cagney wrote:
> My change allows Code Sorcery to achieve their goal of getting Solaris 
> 10 support in GDB, while at the same time allow us to move forward with 
> our objective of improving support for GNU, GNU/Linux and even the other 
> mainstream Free and non-Free platform support.
> 
> We win - Code Sorcery Wins; we have a symbiotic relationship.

First of all, the name of my and Joseph's employer is CodeSourcery. 
The rest of this message is written as a GDB developer, not an
employee.

> On the other hand, by effectively requiring that a contributor must 
> first test/fix a change on marginal if not irrelevant systems such as 
> vax-dec-ultrix4 (the suggestion also carried other less pleasant 
> undertones), can only stall the host's (GDB's) development.  Isn't that 
> called a parasitic relationship?

And you're complaining about Mark's tone?  Please make a passing
effort to be polite.

By requiring contributors to make an architectural change to GDB -
which so far I've seen at least three GDB global maintainers take a
stab at and none finish - you are making GDB more difficult to
contribute to.  This has the effect of driving away contributions,
which isn't any kind of relationship at all.

The timing of deprecating the TM_FILE mechanism was never discussed; it
got lost in your argument with Eli about xm-go32.h.  I apologize for
not loudly objecting at the time; I was making an obviously futile
effort to stay out of an increasingly unpleasant argument.  The tone
of the GDB development lists has gotten steadily worse over the last
year.

I think that deprecating this mechanism with so many unfinished
questions on how to do without it was premature.  The next person to
attempt to contribute a solib-using port obviously got stuck with the
entire mess.

I'll refer to Eli's message:
  http://sources.redhat.com/ml/gdb-patches/2004-09/msg00151.html
The exact same problems apply to TM_FILE.



Separately, the other issue in the above.

Thousands of people use GDB on embedded targets that, Mark said, would
be broken by the patch you posted - targets which do not normally use
shared libraries.  I think it was entirely reasonable of him to insist
that you address his concern before proceeding with surgery on the
solib mechanism.

-- 
Daniel Jacobowitz