Re: Minor PA GAS testsuite fix

[law at redhat dot com]

  > No, what you've done in removing the whitespace before the .import, is
  > turn it into a label.  hppa defines LABELS_WITHOUT_COLONS.  Please put
  > the test back as it was, and remove the xfail.
Somewhat off-topic, but possibly useful to some...

This is an excellent example of the problems associated with coding to
an implementation rather than a specification.  Back in the early 90s
when myself and others spent significant amounts of time bringing GAS
up to snuff on the PA we didn't have any kind of reasonable documentation
on the HP assembler syntax.  The manuals were vague at best and were
missing vast amounts of information.

So we poked and prodded the HP assembler to determine it's behavior and
tried to make GAS emulate it to a reasonable extent -- largely because we
were also stuck trying to assemble a fair amount of hand-written HP
assembler code with gas :(

The result was, IMHO, an assembler that while functional, left a lot to
be desired in the way of long term maintainability and cleanliness.   This
is especially true as the behavior of HP's assembler has changed in 
significant ways over time and I've layered hack upon hack to deal with
the changes.   I'm not all happy with the end result, but I'm not in a
position to go back and revamp things completely.

On a more positive note, we did drop most compatibility with HP's assembler
for PA64!  I don't want to remember the horror of trying to do things like
translate a space/subspace pair into the ELF equivalents.  I did that in
a past life, and it was even grosser than the existing HP support.

Ugh. 

jeff