Re: Regarding run_link_test (was: Re: patch, ld ELF segfault, gasi370-elf, d10v)

[Hans-Peter Nilsson <hp at bitrange dot com>]

On Mon, 10 Jun 2002 trix@redhat.com wrote:
> Hans-Peter Nilsson wrote:
>
> > Why do you add run_link_test?  It's just run_dump_test with no
> > output, right?  If there's some tweak needed, I don't think we
> > want a copy-pasted function with the same basic functionality.
> >
>
> This a design issue.

Not IMO.  This is duplicating code.

> Add a similar test proc or

Mostly *the same* test procedure, with some functionality
stripped off.  So when someone wants a new option, there are now
two places to add it (or three if you count run_dump_test in gas
too; they should merge to some common dir).

> Add another parameter to run_dump_test to circumvent the purpose
> run_dump_test.

No, no parameter needed, and the purpose is not to test a dump.
:-)  The purpose is to test that linking went fine, usually by a
succesful linker exit, absence of error/warning output and then
with some inspection tool.  Alternatively, that the linker emits
a particular error.  So by extension, if you just want to test
that the linker does *not* emit an error (that is, you don't
want to inspect the linked object), why not add a "just-link:"
or "noerror:" or "nothing-here-to-see--move-along:" option?

Though usually, there's *something* about the linked object you
want to inspect, just for sanity's sake.  That's why I've never
found a pressing need for this no-inspection-needed
functionality.

brgds, H-P