[rfa/doco] Test Suite: build != host

[rfa/doco] Test Suite: build != host

[Michael Chastain]

This patch documents that the gdb test suite is supported with
build != host, but that some test scripts do not work and
will return UNRESOLVED if build != host.

I don't know if "UNRESOLVED" needs some kind of @foo{...}
markup.

Testing: rebuilt gdb doco, it looked okay.

Okay to commit?

Michael

===

2004-08-19  Michael Chastain  <mec.gnu@mindspring.com>

	* (Using the Testsuite): build != host is supported,
	but some test scripts do not support build != host.

Index: gdbint.texinfo
===================================================================
RCS file: /cvs/src/src/gdb/doc/gdbint.texinfo,v
retrieving revision 1.219
diff -c -3 -p -r1.219 gdbint.texinfo
*** gdbint.texinfo	14 Aug 2004 21:11:43 -0000	1.219
--- gdbint.texinfo	19 Aug 2004 08:59:19 -0000
*************** difficult to test, such as code that han
*** 6279,6284 ****
--- 6279,6289 ----
  in particular versions of compilers, and it's OK not to try to write
  tests for all of those.
  
+ DejaGNU supports separate build, host, and target machines.  However,
+ some @value{GDBN} test scripts do not work if the host machine is
+ separate from the build machine.  These scripts will return an
+ UNRESOLVED result in such an environment.
+ 
  @section Testsuite Organization
  
  @cindex test suite organization

Re: [rfa/doco] Test Suite: build != host

[Andreas Schwab]

Michael Chastain <mec.gnu@mindspring.com> writes:

> This patch documents that the gdb test suite is supported with
> build != host, but that some test scripts do not work and
> will return UNRESOLVED if build != host.
>
> I don't know if "UNRESOLVED" needs some kind of @foo{...}
> markup.

IMHO @samp would be the right one.

Andreas.

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Maxfeldstraße 5, 90409 Nürnberg, Germany
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: [rfa/doco] Test Suite: build != host

[Eli Zaretskii]

> From: Andreas Schwab <schwab@suse.de>
> Date: Thu, 19 Aug 2004 11:56:33 +0200
> 
> Michael Chastain <mec.gnu@mindspring.com> writes:
> 
> > This patch documents that the gdb test suite is supported with
> > build != host, but that some test scripts do not work and
> > will return UNRESOLVED if build != host.
> >
> > I don't know if "UNRESOLVED" needs some kind of @foo{...}
> > markup.
> 
> IMHO @samp would be the right one.

It depends on what "UNRESOLVED" means, exactly.  I must confess I
don't understand what does it mean "return UNRESOLVED".  Is it a
literal string printed by the test script, or is it something else?

Re: [rfa/doco] Test Suite: build != host

[Michael Chastain]

andreas> IMHO @samp would be the right one.
eli> It depends on what "UNRESOLVED" means, exactly.  I must confess I
eli> don't understand what does it mean "return UNRESOLVED".  Is it a
eli> literal string printed by the test script, or is it something else?

That's right, it's a literal string.

Here is an excerpt from a gdb.sum file:

  Running /berman/fsf/_today_/source/gdb/HEAD/src/gdb/testsuite/gdb.base/unload.exp ...
  gdb compile failed, /berman/fsf/_today_/source/gdb/HEAD/src/gdb/testsuite/gdb.base/unload.c: In function `main':
  /berman/fsf/_today_/source/gdb/HEAD/src/gdb/testsuite/gdb.base/unload.c:39: warning: incompatible implicit declaration of built-in function 'exit'
  /berman/fsf/_today_/source/gdb/HEAD/src/gdb/testsuite/gdb.base/unload.c:47: warning: incompatible implicit declaration of built-in function 'exit'
  WARNING: Testcase compile failed, so all tests in this file will automatically fail.

  ERROR: (timeout) GDB never initialized after 10 seconds.
  WARNING: remote_expect statement without a default case?!
  ERROR: couldn't load /tmp/migbat-testgdb-49vpUkli/test/gdb.base/unload into gdb (end of file).
  UNRESOLVED: gdb.base/unload.exp: set pending breakpoint
  FAIL: gdb.base/unload.exp: single pending breakpoint info
  FAIL: gdb.base/unload.exp: running program
  FAIL: gdb.base/unload.exp: continuing to end of program
  FAIL: gdb.base/unload.exp: rerun to shared library breakpoint
  FAIL: gdb.base/unload.exp: continuing to end of program second time

This script, unload.exp, has some WARNING, ERROR, UNRESOLVED,
and FAIL results.

(The actual bug has already been fixed, I'm just grabbing this
 excerpt to illustrate UNRESOLVED).

My wording is:

  DejaGNU supports separate build, host, and target machines.  However,
  some @value{GDBN} test scripts do not work if the host machine is
  separate from the build machine.  These scripts will return an
  UNRESOLVED result in such an environment.

I'm open to alternate wording, and I'm definitely open to improving
the markup.

Re: [rfa/doco] Test Suite: build != host

[Eli Zaretskii]

[Resending, as it seems like the previous message didn't make it for
some reason.]

> Date: Fri, 20 Aug 2004 11:08:01 -0400
> From: Michael Chastain <mec.gnu@mindspring.com>
> 
> That's right, it's a literal string.
> 
> Here is an excerpt from a gdb.sum file:
> 
>   Running /berman/fsf/_today_/source/gdb/HEAD/src/gdb/testsuite/gdb.base/unload.exp ...
>   gdb compile failed, /berman/fsf/_today_/source/gdb/HEAD/src/gdb/testsuite/gdb.base/unload.c: In function `main':
>   /berman/fsf/_today_/source/gdb/HEAD/src/gdb/testsuite/gdb.base/unload.c:39: warning: incompatible implicit declaration of built-in function 'exit'
>   /berman/fsf/_today_/source/gdb/HEAD/src/gdb/testsuite/gdb.base/unload.c:47: warning: incompatible implicit declaration of built-in function 'exit'
>   WARNING: Testcase compile failed, so all tests in this file will automatically fail.
> 
>   ERROR: (timeout) GDB never initialized after 10 seconds.
>   WARNING: remote_expect statement without a default case?!
>   ERROR: couldn't load /tmp/migbat-testgdb-49vpUkli/test/gdb.base/unload into gdb (end of file).
>   UNRESOLVED: gdb.base/unload.exp: set pending breakpoint

In that case, I suggest to reword the text slightly:

   DejaGNU supports separate build, host, and target machines.  However,
   some @value{GDBN} test scripts do not work if the host machine and
   the build machine are not the same.  These scripts will say that the
   result is ``UNRESOLVED'' in such an environment, like this:

   @smallexample
   UNRESOLVED: gdb.base/unload.exp: set pending breakpoint
   @end smallexample

Re: [rfa/doco] Test Suite: build != host

[Daniel Jacobowitz]

On Thu, Aug 19, 2004 at 05:48:26AM -0400, Michael Chastain wrote:
> This patch documents that the gdb test suite is supported with
> build != host, but that some test scripts do not work and
> will return UNRESOLVED if build != host.
> 
> I don't know if "UNRESOLVED" needs some kind of @foo{...}
> markup.
> 
> Testing: rebuilt gdb doco, it looked okay.
> 
> Okay to commit?

I'd just like to say... one of the few examples you had in favor of
build != host testing was crosstool.  Whoever gave you that example was
mistaken; crosstest.sh only supports host != target, so people are
probably -not- testing gcc with build != host.

I still think this requirement is a bad idea, but it's your call.

-- 
Daniel Jacobowitz

Re: [rfa/doco] Test Suite: build != host

[Michael Chastain]

Daniel Jacobowitz <drow@false.org> wrote:
> I'd just like to say... one of the few examples you had in favor of
> build != host testing was crosstool.  Whoever gave you that example was
> mistaken; crosstest.sh only supports host != target, so people are
> probably -not- testing gcc with build != host.

Rats, you are right.  Or, rather, cool!  You are right!

I got my info from:

  http://kegel.com/crosstool/current/doc/dejagnu-remote-howto.html

And somehow I got the impression that it was using --host.
Which is just plain wrong; it's all about --target_board.

> I still think this requirement is a bad idea, but it's your call.

I'm not happy either way.

I'm working towards a goal where anybody can run the doco; run the test
bed on any configuration supported in the doco; and it always works and
is useful.  Like a real computer program!

One very very hard part is figuring out how to co-operate with all
the contributors and still do wrenching infrastructure upgrades.

Michael