User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:32.0) Gecko/20100101 Firefox/32.0 Build ID: 20140924083558 Steps to reproduce: hg pull -u MOZCONFIG=fennec_mozconfig ./mach build Actual results: 4:10.36 replace_malloc.o 4:10.39 In file included from /home/ignisvulpis/mozilla-central/memory/build/mozmemory_wrap.c:13:0: 4:10.39 /home/ignisvulpis/mozilla-central/memory/build/malloc_decls.h:48:47: error: conflicting types for 'malloc_usable_size' 4:10.39 MALLOC_DECL(malloc_usable_size, size_t, usable_ptr_t) 4:10.39 ^ 4:10.39 In file included from ../../dist/system_wrappers/malloc.h:3:0, 4:10.39 from /usr/local/android-ndk-r10b/platforms/android-9/arch-arm/usr/include/string.h:33, 4:10.39 from ../../dist/system_wrappers/string.h:3, 4:10.39 from /home/ignisvulpis/mozilla-central/memory/build/mozmemory_wrap.c:5: 4:10.39 /usr/local/android-ndk-r10b/platforms/android-9/arch-arm/usr/include/malloc.h:36:15: note: previous declaration of 'malloc_usable_size' was here 4:10.39 extern size_t malloc_usable_size(const void*); 4:10.39 ^ 4:10.39 4:10.39 In the directory /home/ignisvulpis/mozilla-central/obj-arm-linux-androideabi/memory/build 4:10.39 The following command failed to execute properly: 4:10.39 /usr/bin/ccache /usr/local/android-ndk-r10b//toolchains/arm-linux-androideabi-4.8/prebuilt/linux-x86_64/bin/arm-linux-androideabi-gcc -o mozmemory_wrap.o -c -I../../dist/system_wrappers -include /home/ignisvulpis/mozilla-central/config/gcc_hidden.h -DMOZ_MEMORY_IMPL -DMOZ_REPLACE_MALLOC -DAB_CD=en-US -DNO_NSPR_10_SUPPORT -I/home/ignisvulpis/mozilla-central/memory/build -I. -I../../dist/include -I/home/ignisvulpis/mozilla-central/obj-arm-linux-androideabi/dist/include/nspr -I/home/ignisvulpis/mozilla-central/obj-arm-linux-androideabi/dist/include/nss -fPIC -idirafter /usr/local/android-ndk-r10b//platforms/android-9/arch-arm/usr/include -include ../../mozilla-config.h -DMOZILLA_CLIENT -MD -MP -MF .deps/mozmemory_wrap.o.pp -idirafter /usr/local/android-ndk-r10b//platforms/android-9/arch-arm/usr/include -Wall -Wdeclaration-after-statement -Wempty-body -Wpointer-to-int-cast -Wsign-compare -Wtype-limits -Werror=char-subscripts -Werror=comment -Werror=endif-labels -Werror=enum-compare -Werror=ignored-qualifiers -Werror=int-to-pointer-cast -Werror=multichar -Werror=nonnull -Werror=pointer-arith -Werror=pointer-sign -Werror=return-type -Werror=sequence-point -Werror=trigraphs -Werror=unknown-pragmas -Wno-unused -mandroid -fno-short-enums -fno-exceptions -march=armv7-a -mthumb -mfpu=vfp -mfloat-abi=softfp -mno-unaligned-access -std=gnu99 -fgnu89-inline -fno-strict-aliasing -fno-math-errno -pipe -DNDEBUG -DTRIMMED -g -freorder-blocks -Os -fno-reorder-functions -fomit-frame-pointer /home/ignisvulpis/mozilla-central/memory/build/mozmemory_wrap.c 4:10.39 make[5]: *** [mozmemory_wrap.o] Error 1 Expected results: build fennec

Maybe this is interesting to know too: ignisvulpis@namenlos:~/mozilla-central$ cat fennec_mozconfig # Add the correct paths here: ac_add_options --with-android-ndk="/usr/local/android-ndk-r10b/" ac_add_options --with-android-sdk="/usr/local/adt-bundle-linux-x86_64-20140702/sdk/platforms/android-20/" # android options ac_add_options --enable-application=mobile/android ac_add_options --target=arm-linux-androideabi ac_add_options --with-ccache ac_add_options --disable-gstreamer I am on Ubuntu 14.04 LTS.

At first blush, I'm pretty sure we do not use NDK 10b on the buildbots. That doesn't mean it won't work, but it's essentially unsupported. Redirecting to glandium to see if he's seen this or knows what would be relevant.

My guess is that this line is borked: http://dxr.mozilla.org/mozilla-central/source/memory/build/malloc_decls.h#18 Maybe an AC_TRY_COMPILE/AC_TRY_LINK with and without const to define usable_ptr_t right would be best.

I downloaded and installed android-ndk-r9 and get the same error. Axel MOZCONFIG=fennec_mozconfig ./mach build -j1 ... ... 0:10.50 mozmemory_wrap.o 0:10.55 In file included from /home/ignisvulpis/mozilla-central/memory/build/mozmemory_wrap.c:13:0: 0:10.56 /home/ignisvulpis/mozilla-central/memory/build/malloc_decls.h:48:47: error: conflicting types for 'malloc_usable_size' 0:10.56 MALLOC_DECL(malloc_usable_size, size_t, usable_ptr_t) 0:10.56 ^ 0:10.56 In file included from ../../dist/system_wrappers/malloc.h:3:0, 0:10.56 from /usr/local/android-ndk-r9/platforms/android-9/arch-arm/usr/include/string.h:33, 0:10.56 from ../../dist/system_wrappers/string.h:3, 0:10.56 from /home/ignisvulpis/mozilla-central/memory/build/mozmemory_wrap.c:5: 0:10.56 /usr/local/android-ndk-r9/platforms/android-9/arch-arm/usr/include/malloc.h:36:15: note: previous declaration of 'malloc_usable_size' was here 0:10.56 extern size_t malloc_usable_size(const void*); 0:10.56 ^ 0:10.56 0:10.56 In the directory /home/ignisvulpis/mozilla-central/obj-arm-linux-androideabi/memory/build 0:10.56 The following command failed to execute properly: 0:10.56 /usr/bin/ccache /usr/local/android-ndk-r9//toolchains/arm-linux-androideabi-4.8/prebuilt/linux-x86/bin/arm-linux-androideabi-gcc -o mozmemory_wrap.o -c -I../../dist/system_wrappers -include /home/ignisvulpis/mozilla-central/config/gcc_hidden.h -DMOZ_MEMORY_IMPL -DMOZ_REPLACE_MALLOC -DAB_CD=en-US -DNO_NSPR_10_SUPPORT -I/home/ignisvulpis/mozilla-central/memory/build -I. -I../../dist/include -I/home/ignisvulpis/mozilla-central/obj-arm-linux-androideabi/dist/include/nspr -I/home/ignisvulpis/mozilla-central/obj-arm-linux-androideabi/dist/include/nss -fPIC -idirafter /usr/local/android-ndk-r9//platforms/android-9/arch-arm/usr/include -include ../../mozilla-config.h -DMOZILLA_CLIENT -MD -MP -MF .deps/mozmemory_wrap.o.pp -idirafter /usr/local/android-ndk-r9//platforms/android-9/arch-arm/usr/include -Wall -Wdeclaration-after-statement -Wempty-body -Wpointer-to-int-cast -Wsign-compare -Wtype-limits -Werror=char-subscripts -Werror=comment -Werror=endif-labels -Werror=enum-compare -Werror=ignored-qualifiers -Werror=int-to-pointer-cast -Werror=multichar -Werror=nonnull -Werror=pointer-arith -Werror=pointer-sign -Werror=return-type -Werror=sequence-point -Werror=trigraphs -Werror=unknown-pragmas -Wno-unused -mandroid -fno-short-enums -fno-exceptions -march=armv7-a -mthumb -mfpu=vfp -mfloat-abi=softfp -mno-unaligned-access -std=gnu99 -fgnu89-inline -fno-strict-aliasing -fno-math-errno -pipe -DNDEBUG -DTRIMMED -g -freorder-blocks -Os -fno-reorder-functions -fomit-frame-pointer /home/ignisvulpis/mozilla-central/memory/build/mozmemory_wrap.c 0:10.56 make[5]: *** [mozmemory_wrap.o] Error 1 0:10.56 make[4]: *** [memory/build/target] Error 2 0:10.56 make[3]: *** [compile] Error 2 0:10.56 make[2]: *** [default] Error 2 0:10.56 make[1]: *** [realbuild] Error 2 0:10.56 make: *** [build] Error 2 0:10.60 374 compiler warnings present. 0:11.66 ccache (direct) hit rate: 0.0%; (preprocessed) hit rate: 0.0%; miss rate: 100.0% ignisvulpis@namenlos:~/mozilla-central$ cat fennec_mozconfig # Add the correct paths here: ac_add_options --with-android-ndk="/usr/local/android-ndk-r9/" ac_add_options --with-android-sdk="/usr/local/adt-bundle-linux-x86_64-20140702/sdk/platforms/android-20/" # android options ac_add_options --enable-application=mobile/android ac_add_options --target=arm-linux-androideabi ac_add_options --with-ccache ac_add_options --disable-gstreamer ignisvulpis@namenlos:~/mozilla-central$

What I don't understand is that there is already a test for this here: http://mxr.mozilla.org/mozilla-central/source/memory/jemalloc/src/configure.ac#359 I think the test should be moved (or copied) into mozilla-central/configure.in and this line and following http://dxr.mozilla.org/mozilla-central/source/memory/build/malloc_decls.h#18 should be rewritten as somethink like: typedef JEMALLOC_USABLE_SIZE_CONST void * usable_ptr_t;

That ANDROID_VERSION test added in bug 957450 is just probably wrong.

This works for me. host: ubuntu target: android Created a test in configure.in and removed the bogus ifdef

Comment on attachment 8507166 [details] [diff] [review] hg diff > malloc_usable_size_const_ptr.patch glandium: does this look reasonable?

Comment on attachment 8507166 [details] [diff] [review] hg diff > malloc_usable_size_const_ptr.patch Review of attachment 8507166 [details] [diff] [review]: ----------------------------------------------------------------- ::: memory/build/malloc_decls.h @@ +15,4 @@ > > # include "jemalloc_types.h" > > +typedef MALLOC_USABLE_SIZE_CONST_PTR void * usable_ptr_t; I'd rather avoid a configure test for that unless it's _really_ necessary, and I don't think it should be necessary.

Comment on attachment 8507166 [details] [diff] [review] hg diff > malloc_usable_size_const_ptr.patch Review of attachment 8507166 [details] [diff] [review]: ----------------------------------------------------------------- ::: configure.in @@ +3405,5 @@ > +if test "$ac_cv_malloc_usable_size_const_ptr" = yes ; then > + AC_DEFINE(MALLOC_USABLE_SIZE_CONST_PTR, [const]) > +else > + AC_DEFINE(MALLOC_USABLE_SIZE_CONST_PTR,) > +fi IIRC from when I added the test to upstream jemalloc, you do need to test malloc.h exists first, and default to const, not an empty value ::: memory/build/malloc_decls.h @@ +15,4 @@ > > # include "jemalloc_types.h" > > +typedef MALLOC_USABLE_SIZE_CONST_PTR void * usable_ptr_t; Forget my previous comment, I can see what the problem is: they actually changed the old platform headers in the new NDKs... sigh.

(Slightly inaccurate whiteboard tag, but it'll do for now.)

Regarding Comment 9 : I am a big fan of configure tests. Predecessor directives can become wrong like this one which causing the trouble: # if defined(__linux__) && (!defined(MOZ_MEMORY_ANDROID) || ANDROID_VERSION < 19) configure tests are kind of more authoritative than preprocessor #if's and even with a big number of configure tests the developement setup time is IMO short. Q: shouldn't the status be "NEW" instead of "UNCONFIRMED"?

(In reply to Nick Alexander :nalexander from comment #8) > Comment on attachment 8507166 [details] [diff] [review] > hg diff > malloc_usable_size_const_ptr.patch fwiw this patch helped for Fennec builds, but fails on a Flame b2g build.

(In reply to Douglas Crosher [:dougc] from comment #13) > (In reply to Nick Alexander :nalexander from comment #8) > > Comment on attachment 8507166 [details] [diff] [review] > > hg diff > malloc_usable_size_const_ptr.patch > > fwiw this patch helped for Fennec builds, but fails on a Flame b2g build. Hm, I just build b2g for the nexus-5 with this patch and it built. I added the new test to B2G/gecko/configure.in and changed B2G/gecko/memory/build/malloc_decls.h Why does it fail for you? I would like to add a B2G patch but I do not know how to create one. ignisvulpis@namenlos:~/development/B2G$ git diff -U8 --patience -M -C returns nothing... Strange.

(In reply to Axel Nennker from comment #14) > (In reply to Douglas Crosher [:dougc] from comment #13) > > (In reply to Nick Alexander :nalexander from comment #8) > > > Comment on attachment 8507166 [details] [diff] [review] > > > hg diff > malloc_usable_size_const_ptr.patch > > > > fwiw this patch helped for Fennec builds, but fails on a Flame b2g build. > > Hm, I just build b2g for the nexus-5 with this patch and it built. > I added the new test to B2G/gecko/configure.in and changed > B2G/gecko/memory/build/malloc_decls.h > Why does it fail for you? > > I would like to add a B2G patch but I do not know how to create one. > ignisvulpis@namenlos:~/development/B2G$ git diff -U8 --patience -M -C > returns nothing... Strange. Regarding b2g: I guessed how the diff works here ignisvulpis@namenlos:~/development/B2G$ git diff ignisvulpis@namenlos:~/development/B2G$ hg diff abort: no repository found in '/home/ignisvulpis/development/B2G' (.hg not found)! ignisvulpis@namenlos:~/development/B2G$ ./repo diff project gecko/ diff --git a/configure.in b/configure.in index bc4d9e2..2ffde4e 100644 --- a/configure.in +++ b/configure.in @@ -3420,6 +3420,20 @@ done AC_CHECK_FUNCS(strndup posix_memalign memalign valloc) +AC_CACHE_CHECK(for malloc_usable_size_const_ptr, + ac_cv_malloc_usable_size_const_ptr, + [AC_TRY_COMPILE([#include <malloc.h> + #include <stddef.h> + size_t malloc_usable_size(const void *ptr);], + [], + ac_cv_malloc_usable_size_const_ptr=yes, + ac_cv_malloc_usable_size_const_ptr=no)]) +if test "$ac_cv_malloc_usable_size_const_ptr" = yes ; then + AC_DEFINE(MALLOC_USABLE_SIZE_CONST_PTR, [const]) +else + AC_DEFINE(MALLOC_USABLE_SIZE_CONST_PTR,) +fi + AC_CHECK_FUNCS(malloc_usable_size) dnl See if compiler supports some gcc-style attributes diff --git a/memory/build/malloc_decls.h b/memory/build/malloc_decls.h index e46cf59..ff33ec0 100644 --- a/memory/build/malloc_decls.h +++ b/memory/build/malloc_decls.h @@ -15,11 +15,7 @@ # include "jemalloc_types.h" -# if defined(__linux__) && (!defined(MOZ_MEMORY_ANDROID) || ANDROID_VERSION < 19) -typedef void * usable_ptr_t; -# else -typedef const void * usable_ptr_t; -# endif +typedef MALLOC_USABLE_SIZE_CONST_PTR void * usable_ptr_t; # define MALLOC_FUNCS_MALLOC 1 # define MALLOC_FUNCS_JEMALLOC 2 ignisvulpis@namenlos:~/development/B2G$ I think that the patch works with b2g too.

Silence on this bug for a week... Could someone please set the status from unconfirmed to new? What about the claim that the patch (configure.in) is responsible for B2G bot build failures? I can build B2G on Ubuntu64 with the patch in place. Is there a plan to update B2G/ndk to the latest ndk?

Comment on attachment 8544911 [details] [diff] [review] malloc_usable_size.patch Review of attachment 8544911 [details] [diff] [review]: ----------------------------------------------------------------- ::: configure.in @@ +3405,5 @@ > AC_CHECK_FUNCS(strndup posix_memalign memalign) > > AC_CHECK_FUNCS(malloc_usable_size) > +MALLOC_USABLE_SIZE_CONST_PTR=const > +AC_CHECK_HEADERS([malloc.h], [ MOZ_CHECK_HEADERS @@ +3411,5 @@ > + AC_TRY_COMPILE([#include <malloc.h> > + #include <stddef.h> > + size_t malloc_usable_size(const void *ptr);], > + [return malloc_usable_size(0);], > + , AC_MSG_RESULT([yes]), @@ +3412,5 @@ > + #include <stddef.h> > + size_t malloc_usable_size(const void *ptr);], > + [return malloc_usable_size(0);], > + , > + MALLOC_USABLE_SIZE_CONST_PTR=) AC_MSG_RESULT([no]) @@ +3414,5 @@ > + [return malloc_usable_size(0);], > + , > + MALLOC_USABLE_SIZE_CONST_PTR=) > +]) > +AC_DEFINE_UNQUOTED(MALLOC_USABLE_SIZE_CONST_PTR, $MALLOC_USABLE_SIZE_CONST_PTR) AC_DEFINE_UNQUOTED([MALLOC_USABLE_SIZE_CONST_PTR], [$MALLOC_USABLE_SIZE_CONST_PTR])

sorry had to back this out for bustage like https://treeherder.mozilla.org/logviewer.html#?job_id=5157727&repo=mozilla-inbound

https://hg.mozilla.org/projects/cypress/rev/3d587e85f2a2 https://hg.mozilla.org/projects/cypress/rev/127c5aadbddd

Comment on attachment 8545635 [details] [diff] [review] malloc_usable-win.patch Review of attachment 8545635 [details] [diff] [review]: ----------------------------------------------------------------- ::: memory/build/malloc_decls.h @@ +15,4 @@ > > # include "jemalloc_types.h" > > +#ifdef MALLOC_USABLE_SIZE_CONST_PTR I'd rather go with #ifndef MALLOC_USABLE_SIZE_CONST_PTR #define MALLOC_USABLE_SIZE_CONST_PTR const #endif or move the configure test in a zone where windows builds run it (at least the AC_DEFINE_UNQUOTED and the default value, but it's better if the block stays together)