417345 - build broken on sparc

Status: RESOLVED FIXED

Description

[Fabien Tassin]

1.9b3 build is broken on sparc:

g++-4.2 -o nsCookieService.o -c -fvisibility=hidden -DMOZILLA_INTERNAL_API -D_IMPL_NS_COM -DEXPORT_XPT_API -DEXPORT_XPTC_API -D_IMPL_NS_COM_OBSOLETE -D_IMPL_NS_GFX -D_IMPL_NS_WIDGET -DIMPL_XREAPI -DIMPL_NS_NET -DIMPL_THEBES  -DOSTYPE=\"Linux2.6\" -DOSARCH=Linux -DIMPL_NS_NET  -I. -I. -I../../../dist/include/xpcom -I../../../dist/include/string -I../../../dist/include/pref -I../../../dist/include/storage -I../../../dist/include   -I../../../dist/include/necko -I/usr/include/nspr -I/usr/include  -I/usr/include  -I../../../dist/sdk/include    -fPIC   -fno-rtti -fno-exceptions -Wall -Wconversion -Wpointer-arith -Woverloaded-virtual -Wsynth -Wno-ctor-dtor-privacy -Wno-non-virtual-dtor -Wcast-align -Wno-long-long -pedantic -g -fno-strict-aliasing -fshort-wchar -pthread -pipe  -DNDEBUG -DTRIMMED -Os -freorder-blocks -fno-reorder-functions -Werror   -DMOZILLA_CLIENT -include ../../../mozilla-config.h -Wp,-MD,.deps/nsCookieService.pp nsCookieService.cpp
cc1plus: warnings being treated as errors
In file included from ../../../dist/include/xpcom/nsCOMArray.h:42,
                 from nsCookieService.cpp:59:
../../../dist/include/xpcom/nsVoidArray.h: In member function 'void nsAutoVoidArray::ResetToAutoBuffer()':
../../../dist/include/xpcom/nsVoidArray.h:193: warning: cast from 'char*' to 'nsVoidArray::Impl*' increases required alignment of target type
make[6]: *** [nsCookieService.o] Error 1
make[6]: Leaving directory `/build/buildd/xulrunner-1.9~b3/build-tree/mozilla/netwerk/cookie/src'

Comment 1

[[e-mail]]

Created attachment 305601 [details] [diff] [review]
fix nsVoidArray

does this fix your build?

Comment 2

[[e-mail]]

note, dbaron has a patch in bug 409384 that should also fix this for nsVoidArray.

Comment 3

[Dennis Gilmore]

it does not fix it for me.  I am building with gcc-4.3  build logs can be found http://sparc.koji.fedoraproject.org/koji/taskinfo?taskID=51185

Comment 4

[[e-mail]]

ok, i'm going to take another stab at a fix (for real, this time!).

in the meantime, this isn't a cookie bug, -> core/xpcom

Comment 5

[[e-mail]]

Created attachment 309157 [details] [diff] [review]
v2

turn off -Wcast-align on sparc. let it join ia64 in its uniqueness.

Comment 6

[[e-mail]]

(i tried fixing nsTHashtable, but it's not possible - we know the casts are correct but there's no way of telling the compiler so, short of sprinkling __align() constructs. other option is to disable WARNINGS_AS_ERRORS for cookies on sparc, but this is a pretty pervasive warning and sparc is a long shot from tier 1.)

Comment 7

[[e-mail]]

Comment on attachment 309157 [details] [diff] [review]
v2

requesting approval for a trivial build fix on sparc.

Comment 8

[Damon Sicore (:damons)]

Comment on attachment 309157 [details] [diff] [review]
v2

a1.9+=damons

Comment 9

[[e-mail]]

Checking in configure.in;
/cvsroot/mozilla/configure.in,v  <--  configure.in
new revision: 1.1955; previous revision: 1.1954
done

please reopen if this doesn't fix things for you guys. ;)

Comment 10

[Dennis Gilmore]

Created attachment 309348 [details] [diff] [review]
Patch Fixing for sparc64 

the previous patch fixed 32bit sparc  but not 64bit sparc  the attached patch fixed that part of the build for me.  I hit a different failure that i will open a new bug for

Comment 11

[Dennis Gilmore]

I am unable to reopen the bug  please do so 

Comment 12

[Raúl Porcel]

Reopening :)

Comment 13

[Ted Mielczarek [:ted.mielczarek]]

Comment on attachment 309348 [details] [diff] [review]
Patch Fixing for sparc64 

Could you just change this to a case statement at this point? It's starting to get ugly.

Comment 14

[[e-mail]]

dennis, want to roll a new patch and request review from :luser?

Comment 15

[Dennis Gilmore]

Dan Can do

Comment 16

[Dennis Gilmore]

Created attachment 310881 [details] [diff] [review]
rewritten patch using case statement

Comment 17

[[e-mail]]

Comment on attachment 310881 [details] [diff] [review]
rewritten patch using case statement

>+    case "$CPU_ARCH" in
>+    ia64 | sparc | sparc64)
>+        # don't use -Wcast-align on ia64, sparc or sparc64, it's noisy on those platforms
>         _WARNINGS_CFLAGS="${_WARNINGS_CFLAGS} -Wcast-align"

looks like you've got your sense mixed up... you want the _WARNINGS_CFLAGS bit to be in the * case, no? ;)

Comment 18

[Dennis Gilmore]

Created attachment 310889 [details] [diff] [review]
Really update to use case statement 

you were right.  I was still building with the patch appalied.  the new one is better

Comment 19

[Dennis Gilmore]

it builds on sparc32 with the patch applied.  testing sparc64 now.

Comment 20

[Dennis Gilmore]

Created attachment 310968 [details] [diff] [review]
make sparc same as ppc

The patch doesnt apply to current cvs checkouts due to adding to the if statement to handle the intel compiler. i refactored handing of sparc cpus to be the same as powerpc.  ive built on sparc64 with the new patch applied.

Comment 21

[[e-mail]]

Comment on attachment 310968 [details] [diff] [review]
make sparc same as ppc

trivial build fix for sparc64.

Comment 22

[Mike Beltzner [:beltzner, not reading bugmail]]

Comment on attachment 310968 [details] [diff] [review]
make sparc same as ppc

a=beltzner

Comment 23

[Reed Loden [:reed] (use needinfo?)]

Checking in configure.in;
/cvsroot/mozilla/configure.in,v  <--  configure.in
new revision: 1.1969; previous revision: 1.1968
done