1064665 - no makefile in intl/icu/target

Status: NEW

(Firefox Build System :: General, defect)

Description

[Landry Breuil (:gaston)]

I've seen it seemingly random since a while, but 100% reproducible when building 33.0b1 - fails in icu, with an empty intl/icu/target dir. Dunno where/when a makefile should be generated there, but i have nothing...

gmake[4]: Entering directory '/usr/obj/ports/firefox-33.0beta1/build-amd64/intl/icu'
gmake[4]: Nothing to be done for 'tools'.
gmake[4]: Leaving directory '/usr/obj/ports/firefox-33.0beta1/build-amd64/intl/icu'
if test -d ../../dist/bin/browser ; then touch ../../dist/bin/browser/.purgecaches ; fi
if test -d ../../dist/bin/webapprt ; then touch ../../dist/bin/webapprt/.purgecaches ; fi
gmake -j1 -C target STATIC_O=o GENRBOPTS='-kR'
gmake[4]: Entering directory '/usr/obj/ports/firefox-33.0beta1/build-amd64/intl/icu/target'
gmake[4]: *** No targets specified and no makefile found.  Stop.
gmake[4]: Leaving directory '/usr/obj/ports/firefox-33.0beta1/build-amd64/intl/icu/target'
Makefile:66: recipe for target 'buildicu' failed
gmake[3]: *** [buildicu] Error 2

any hints welcome

Comment 1

[Landry Breuil (:gaston)]

After digging a bit more, it seems running configure within 33.0b1 doesnt run configure for intl/icu, it directly goes to libffi.

m-c:

 1:46.30 updating cache ./config.cache
 1:46.31 creating ./config.status
 1:46.66 configuring in intl/icu/target
 1:46.66 running /bin/sh /src/mozilla-central/intl/icu/source/configure --build=i386-unknown-openbsd5.6 --host=i386-unknown-openbsd5.6 --en
able-static --disable-shared --disable-extras --disable-icuio --disable-layout --disable-tests --disable-samples --cache-file=/usr/obj/tmpf
s/m-c/intl/icu/target/config.cache
 1:46.86 configure: creating cache /usr/obj/tmpfs/m-c/intl/icu/target/config.cache


33.0b1:

updating cache ./config.cache
creating ./config.status
configuring in js/src/ctypes/libffi

i'm... puzzled.

Comment 2

[Michael Shal [:mshal]]

Maybe it's accidentally getting skipped because of bug 1043268? You could try backing out that patch (and maybe the followup patch too) to see if it's related.

Comment 3

[Landry Breuil (:gaston)]

the intl api is enabled and exposed according to config.status, i dont use with-system-icu...

after adding some debug echos to configure, it seems a subshell is supposed to be done to run the subconfigure, but the closing parenthesis is missing.. so maybe an autoconf issue.

29629         (export AR="$HOST_AR"
29630          export RANLIB="$HOST_RANLIB"
29631          export CC="$HOST_CC"
29632          export CXX="$HOST_CXX"
29633          export CPP="$HOST_CPP"
29634          export LD="$HOST_LD"
29635          export CFLAGS="$HOST_ICU_CFLAGS $HOST_OPTIMIZE_FLAGS"
29636          export CPPFLAGS="$ICU_CPPFLAGS"
29637          export CXXFLAGS="$HOST_ICU_CXXFLAGS $HOST_OPTIMIZE_FLAGS"
29638          export LDFLAGS="$HOST_LDFLAGS"
29639          ac_configure_args="$HOST_ICU_BUILD_OPTS"
29640          ac_configure_args="$ac_configure_args --enable-static --disable-shared --enable-extras=no --enable-icuio=no --enable-layout=      no --enable-tests=no --enable-samples=no"
29641          trap '' EXIT
29642 for moz_config_dir in intl/icu/source:intl/icu/host; do
29643   case "$moz_config_dir" in
29644   *:*)
*** 300 more lines then ***
30021 trap '[ "$?" != 0 ] && echo "------ config.log ------" && tail -n 25 config.log' EXIT
30022 
30023 
30024         ) || exit 1
30025     fi


Again, i'm suspecting a gnu m4 vs bsd m4 issue... since in build/autoconf/icu.m4, theres:

180         (export AR="$HOST_AR"
181          export RANLIB="$HOST_RANLIB"
182          export CC="$HOST_CC"
183          export CXX="$HOST_CXX"
184          export CPP="$HOST_CPP"
185          export LD="$HOST_LD"
186          export CFLAGS="$HOST_ICU_CFLAGS $HOST_OPTIMIZE_FLAGS"
187          export CPPFLAGS="$ICU_CPPFLAGS"
188          export CXXFLAGS="$HOST_ICU_CXXFLAGS $HOST_OPTIMIZE_FLAGS"
189          export LDFLAGS="$HOST_LDFLAGS"
190          ac_configure_args="$HOST_ICU_BUILD_OPTS"
191          ac_configure_args="$ac_configure_args --enable-static --disable-shared --enable-extras=no --enable-icuio=no --enable-layout=no     --enable-tests=no --enable-samples=no"
192          AC_OUTPUT_SUBDIRS(intl/icu/source:intl/icu/host)
193         ) || exit 1

i suspect AC_OUTPUT_SUBDIRS to behave strangely here in my infrastructure. Anyway, i suspect its a fallout from 1036645, mike, correct me if my analysis looks wrong.

Comment 4

[Landry Breuil (:gaston)]

(In reply to Michael Shal [:mshal] from comment #2)
> Maybe it's accidentally getting skipped because of bug 1043268? You could
> try backing out that patch (and maybe the followup patch too) to see if it's
> related.

i doubt that, since i'm starting from an empty objdir, there should be no reason for ignoring running subconfigures.. and bug 1043268 landed for 34.