Within my duties as Solaris GCC maintainer, I'm currently testing GCC
mainline with several different toolchains: Sun as and ld, GNU as and
Sun ld, and GNU as and ld, to check for feature parity and bugs in the
various combinations.
During that work, I found that the assembler documentation is seriously
lacking:
* The SPARC Assembly Language Reference Manual was last updated in 2002,
the x86 Assembly Language Reference Manual in 2005.
* The important information to anyone writing assembler input (either
directly or in a compiler) are the details of the assembler directives
understood. This section is both incomplete (understandable given the
age of the manuals, e.g. there's no information on the .group
directive) and lacks syntactic and semantic detail necessary to use
it.
* Assembler error messages are not explained at all; some of them are
unfortunately all but self-explaining.
All this caused me to either try and discover missing information by
trial and error, looking at Studio cc/CC -S output, and perhaps looking at
the Solaris 9 as sources from the edu source program, though they are
quite dated by now.
It would be very good to remedy this situation: I consider the Linker
and Libraries Guide to be exemplary in this regard: it contains both
conceptual/high-level information on the concepts, lots of examples, and
reference chapters with all the details necessary to understand and
dissect object files. Bryan Cantrill once mentioned to me that they
strived to reach that standard with the DTrace Manual :-)
I've no idea if there is any documentation inside Sun that could be used
to improve the existing manuals. Perhaps it's conceivable that the
as/fbe sources can be open-sourced, so at least one could look at the
source to discover missing information.
In the meantime, I've filed a couple of bugs/RFEs at bugs.sun.com, with
more to come. I suppose this is the best place to do so,
bugs.opensolaris.org might be an alternative, though as/fbe bugs won't
be visible there. Unfortunately, none of them have been acknowledged
yet, and I'm reluctant to file more until I can be assured that they are
not lost completely.
Comments?
Thanks,
Rainer
--
-----------------------------------------------------------------------------
Rainer Orth, Center for Biotechnology, Bielefeld University
