Re: object/load module formats

Shaun Wilkinson <shaun@rdc31.std.com> writes:
>We`re thinking of changing the object/load module format of our cross
>tools. Does someone know about standard/popular object/load module format
>which are suitable for our purpose? (we need their documents, license
>free.) What are the current trends in this area?


Our moderator replies:


>[SVR4 has ELF, OSF/1 has ROSE, and GNU has a souped up BSD format. Is anyone
>moving toward anything else? -John]


Ron Guilmette comments:
| Actually, it is my understanding that OSF finally dropped ROSE and decided
| to just go with the (ELF) flow. (I have mixed feeling about this because
| I'm a big proponent of ELF and DWARF, but my good friend Mike Meissner was
| the main instigator of ROSE, and I believe that ROSE had a lot of good
| thought put into its design.)


Umm, I was not the main instigator. Of all the 5 people in the group at
the time, I probably had the least number of changes that went into
what became OSF/rose. The main architect of OSF/rose was Melanie Weaver.


| All things considered, I wouldn't mind a bit if ELF became *the* standard,
| and if *everybody* decided to use it. That sure as hell would cut down on
| a lot of the ``wheel reinvention'' that tools people currently have to do
| when porting from system to system.


ELF has its warts too. The contortions that one has to go through to
get the last text page and first data page to be shared in the file
come to mind as a wart. These contortions are more than just the
compiler suite, and extend into the host OS. The differences between
sections and the program table is another. Not planning for 64 bit
machines and leaving appropritate holes in the structure for the 64
bit sizes to just fall in is a third.


Also the work (phone tag, etc. not actual coding) involved to actually
be able to use ELF legally without a SVR4 license was a nightmare.


--
Michael Meissner email: meissner@osf.org phone: 617-621-8861
Open Software Foundation, 11 Cambridge Center, Cambridge, MA, 02142
--

Return to the comp.compilers page.
Search the comp.compilers archives again.