[Mono-bugs] [Bug 324134] Add support for Mac OS X / ppc64
bugzilla_noreply at novell.com
bugzilla_noreply at novell.com
Sat May 10 18:42:58 EDT 2008
User andreas.faerber at web.de added comment
--- Comment #9 from Andreas Färber <andreas.faerber at web.de> 2008-05-10 16:42:55 MST ---
Thanks for your comments, Geoff! Unfortunately they are a little vague: (and it
appears you're not fully aware of its history)
The only whitespace changes I'm aware of are in ppc-codegen.h, and as stated
above, they are waiting for a separate reformatting-only patch, which hasn't
been reviewed yet (it was a response to Paolo saying that, too). Quote: "Since
it has not been reviewed, the real patch is not yet rebased against it." Once
agreed on how to layout the file, my working copy diff can more sensibly be
updated to only include contential changes in that file. (cf. comment #6)
CHECKME obviously indicates that that place needs to be verified for
correctness or for necessity of additional changes; it is not intended to go
into SVN. This is not a complete patch (it says "draft patch" or working copy
diff), so you cannot just look over it and say "okay to commit" - it might well
be that things need fixing beyond this patch that you would only see by
applying the diff to a clean working copy (no problem with whitespace then
either). If you know for sure that some CHECKME can be removed, let me know.
No idea what you mean with codepaths between sizes and warnings. The signedness
warnings I fixed do apply to both codepaths.
Where is such a define?
configure.in: changes for eglib were necessary to link, specifically, -liconv
was needed for OSX but not on Linux; the .la file is the proper solution to
take care of such differences (part one). If I leave out the second part,
monoburg does not link due to lack of -liconv or (with part one) because gcc
can't handle .la file. If you have a better suggestion, please propose. (it's
not just about OSX, a few platforms don't even allow hardcoding -lm there) The
previous version of these changes was originally for v10.4, but v10.5 appeared
to have the same issues on first try. Some configure.in changes had already
been committed from a separate WC in r102509; it is included in this patch by
nature of `svn diff` and because it is necessary to try out the port in lack of
a git or SVN branch.
exceptions-ppc.c: No idea what you mean... if in fact there is such a thing,
it's probably not intended.
tramp_ppc.c: sp == r1. Made the code better readable for me in my working copy;
code switched between the two and should be a little more consistent in my diff
- I'm pretty sure I didn't change ppc_sp to ppc_r1 anywhere, I changed ppc_r1
to ppc_sp (sp == Stack Pointer)
Among all the various things in this working copy, ppc-codegen.h really is the
key change. Getting that into shape is the basis for all mini/ changes and thus
needs to be addressed first. This involves reviewing...
a) the move of copyright info (anything still wrong/missing/TODO?),
b) the reordering of some instruction macros (I proposed alphabetically for
instructions, and by instructions for mnemonics),
c) the code style (most existing ppc macros violate the Mono coding guidelines;
not fully addressed in the reformatting patch since it would make it
d) the insertion of newlines for grouping (as opposed to indenting in WC diff)
This can then obviously be further split up into multiple patches, if
Will be off over the holidays myself.
Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.
More information about the mono-bugs