This Makefile.in contains a variety of horrors: https://hg.mozilla.org/integration/mozilla-inbound/file/33d36bf6ca0c9d9c22cadf6d8223fa6e1418b62c/xpcom/reflect/xptcall/md/unix/Makefile.in It looks like they're basically all for non-Tier-1 platforms, though, and primarily platforms we don't even really care about supporting anymore. We should probably just remove the Makefile.in and if it breaks someone on a Tier-3 platform they can fix it. The only things that someone might actually care about are: * glandium's hppa workarounds from bug 434190 * The OS X PPC assembly generation, xptcstubs_asm_ppc_darwin.s, which uses m4(!) to generate an assembly file. The Mac PPC maintainers can fix that, the simplest solution would be to just stick it in GENERATED_FILES and make the Python script invoke m4. (Still horrible, but at least not in a Makefile.in). I'm 100% positive we don't care about HP-UX or AIX, and we certainly don't care enough about Solaris to do the work to port the Makefile bits there.

(In reply to Ted Mielczarek [:ted.mielczarek] from comment #0) > * The OS X PPC assembly generation, xptcstubs_asm_ppc_darwin.s, which uses > m4(!) to generate an assembly file. The Mac PPC maintainers can fix that, > the simplest solution would be to just stick it in GENERATED_FILES and make > the Python script invoke m4. (Still horrible, but at least not in a > Makefile.in). Which "Python script" are you referring to? Internally I have something in our changesets related to this, which I assume is the same thing you mean: if CONFIG['OS_ARCH'] == 'Darwin': SOURCES += [ 'xptcinvoke_asm_ppc_rhapsody.s', 'xptcinvoke_darwin.cpp', 'xptcstubs_darwin.cpp', ] GENERATED_SOURCES += [ 'xptcstubs_asm_ppc_darwin.S', ] if CONFIG['OS_TEST'] == 'powerpc': SOURCES += [ 'xptcinvoke_asm_ppc_rhapsody.s', ] GENERATED_SOURCES += [ 'xptcstubs_asm_ppc_darwin.S', ] if '86' in CONFIG['OS_TEST'] and CONFIG['OS_TEST'] != 'x86_64': DEFINES['MOZ_NEED_LEADING_UNDERSCORE'] = True > I'm 100% positive we don't care about HP-UX or AIX, and we certainly don't > care enough about Solaris to do the work to port the Makefile bits there. I don't know if Ginn still reads bugmail, but someone is still making contrib builds for Solaris 10. I'm using 38ESR on my Ultra-3 SPARC laptop and it works pretty well, considering.

(added Ginn for comments on the Solaris piece)

(In reply to Cameron Kaiser [:spectre] from comment #1) > (In reply to Ted Mielczarek [:ted.mielczarek] from comment #0) > > * The OS X PPC assembly generation, xptcstubs_asm_ppc_darwin.s, which uses > > m4(!) to generate an assembly file. The Mac PPC maintainers can fix that, > > the simplest solution would be to just stick it in GENERATED_FILES and make > > the Python script invoke m4. (Still horrible, but at least not in a > > Makefile.in). > > Which "Python script" are you referring to? Sorry, I should have clarified that, I meant "a Python script could be written to be invoked by GENERATED_FILES", using the .script flag: http://gecko.readthedocs.org/en/latest/build/buildsystem/mozbuild-symbols.html#generated-files > Internally I have something in our changesets related to this, which I > assume is the same thing you mean: That's part of it, declaring the file as a GENERATED_FILE. The other half is moving the generation out of a Makefile rule into a Python script. In this case you could just make the Python script invoke the ridiculous stuff the existing Makefile rules do. > I don't know if Ginn still reads bugmail, but someone is still making > contrib builds for Solaris 10. I'm using 38ESR on my Ultra-3 SPARC laptop > and it works pretty well, considering. If he wants to fix it that's fine, we're just not going to spend time on tricky edge cases for platforms like Solaris.

Comment on attachment 8915079 [details] bug 1253607 - remove xpcom/reflect/xptcall/md/unix/Makefile.in. https://reviewboard.mozilla.org/r/186340/#review191528 Should we be removing the corresponding moz.build stuff for these platforms as well? It's a little odd to have !xptcstubs_asm_ppc_darwin.s in the moz.build file but no way to generate it: https://dxr.mozilla.org/mozilla-central/rev/c97190c389c4cfef20fe55b4bacade95a36ae6ef/xpcom/reflect/xptcall/md/unix/moz.build#14

Oh, yeah, I guess that would be a sensible thing to do. :) In my defense I generated this patch at like 6:30AM.

After looking at the moz.build file, I have decided I'm going to land this simple removal patch as-is and file a followup for cleanup. There's a lot of stuff that can be removed from the moz.build file for platforms we no longer support (Mac PPC, AIX, etc) and I don't want to overcomplicate this patch.

https://hg.mozilla.org/integration/mozilla-inbound/rev/c801b23bbf1a434a958b5db5a48bed772b5af2e3 bug 1253607 - remove xpcom/reflect/xptcall/md/unix/Makefile.in. r=mshal