Only I use gas under x86?

Only I use gas under x86?

[H.J. Lu]

Hi,

After installing the new gas, my old programs assembled by the new gas
failed to work. I spent a whole night tracking it down. It turned out
the new x86 gas was broken due to a typo. Please tak a second look at
diffs-950411.gz and pay attenyion to ./gas/config/tc-i386.h. BTW,
there was no ChangeLog entry for the typo.

No, I won't provide a diff nor point out the typo. It is a very dumb
typo.

I have a feeling that not many people test the x86 gas a lot.


-- 
H.J. Lu
NYNEX Science and Technology, Inc.			hjl@nynexst.com

Re: Only I use gas under x86?

[Roland McGrath]

I use gas on i386 all the time (target=i386-gnuelf in fact, which is
identical to i386-linux for gas).  I use the current sources, though I
don't recompile every day.  Whenever there has been a problem it has been
taken care of pretty quickly.  We cannot expect the maintainers to test
every single change on every platform; we are using development code,
certainly it would be tested on i386 before released.

However, it is a serious problem if people are making changes without
proper ChangeLog entries.  That really makes life difficult.  The rest of
us do not have access to the RCS logs to fill in the gaps in the ChangeLog
files; those files really need to be complete.  They are the first place
anyone looks to see what changed when something broke, and why.

Re: Only I use gas under x86?

[Richard Stallman]

RCS log items should be copies of material from ChangeLog.
People installing changes in GAS should first write entries
in ChangeLog, then copy them for RCS checkin.

(There is one other alternative--to generate ChangeLog from the RCS
logs--but this is an all-or-nothing choice.  If the maintainers of a
program switch to that method, they must all switch at once.)

Re: Only I use gas under x86?

[Ken Raeburn]

	 BTW,
   there was no ChangeLog entry for the typo.

Actually, there was -- two change log entries for changes to
practically all the tc-*.c or .h files.

   No, I won't provide a diff nor point out the typo. It is a very dumb
   typo.

Yup, it was.

   I have a feeling that not many people test the x86 gas a lot.

My automated tests for x86 are still broken because (I think) someone
deinstalled bison from some of the Cygnus machines.  And we've only
got two in the California office I can do the automatic testing on,
running sco and unixware.  It's not a happy scene.

Once I get a few "I need a patch last week" customer bug reports out
of the way, I plan to clean up some of the problems reported in the
automatic testing.  And maybe even start catching up in my
bug-gnu-utils backlog. :-(

For the moment, yes, I think you're one of the main x86 testers of the
development code.

Maybe I'll look into some automatic testing in the Massachusetts
office; we've got a Linux box there, but network security issues make
it a little difficult to tie the testing into the snapshot process as
tightly as I'd like.  Tying in my home pc (netbsd) or laptop (linux)
would be at least as difficult.