On Thu, Oct 25, 2018 at 01:08:36PM -0500, Bruce Dubbs via blfs-dev wrote:
On 10/25/2018 09:31 AM, renodr via blfs-dev wrote:
Here's a list of my packages (at least as it pertains to Firefox
dependencies):
Wed Oct 17 22:26:59 CDT 2018 autoconf-2.13
Tue Oct 23 21:02:49 CDT 2018 cbindgen-0.6.6
Tue Oct 16 15:33:12 CDT 2018 llvm-7.0.0
I only have -6.0.1 right now
Hi Bruce,
I guess you'll have seen my post by now.
Thu Oct 18 11:59:38 CDT 2018 gtk+-3.24.1 *
I only have -3.22.30 right now
Should be ok
Tue Oct 16 19:10:44 CDT 2018 icu4c-62.1
I have -63_1
^^^^^ - see the end of my reply
Tue Oct 16 18:19:47 CDT 2018 libevent-2.1.8
Sun Oct 21 19:28:33 CDT 2018 libvpx-1.7.0
Tue Oct 23 21:48:02 CDT 2018 node-v9.11.2
Mon Oct 22 21:45:18 CDT 2018 sqlite-autoconf-3250200
I have -3240000
Probably ok, but sqlite, nss and the certificates are things I
always update for a new firefox release.
Tue Oct 16 16:25:56 CDT 2018 valgrind-3.13.0
I have -3.14.0
I guess that is unlikely to matter
Wed Oct 17 19:50:42 CDT 2018 graphite2-1.3.11
I have -1.3.12
Necessary if enabling system graphite (unless I've botched the
patch). :
+dnl Check for graphite2
+dnl ========================================================
+if test -n "$MOZ_SYSTEM_GRAPHITE2"; then
+ dnl graphite2.pc has bogus version, check manually
+ _SAVE_CFLAGS=$CFLAGS
+ CFLAGS="$CFLAGS $MOZ_GRAPHITE2_CFLAGS"
+ AC_TRY_COMPILE([ #include <graphite2/Font.h>
+ #define GR2_VERSION_REQUIRE(major,minor,bugfix) \
+ ( GR2_VERSION_MAJOR * 10000 + GR2_VERSION_MINOR \
+ * 100 + GR2_VERSION_BUGFIX >= \
+ (major) * 10000 + (minor) * 100 + (bugfix) )
+ ], [
+ #if !GR2_VERSION_REQUIRE(1,3,12)
+ #error "Insufficient graphite2 version."
+ #endif
+ ], [],
+ [AC_MSG_ERROR([--with-system-graphite2 requested but no
working libgraphite2 found])])
+ CFLAGS=$_SAVE_CFLAGS
+fi
That looks as if it should #error with 1.3.11. The shipped version
is 1.3.12. But without enabling it, should work (I only test with
both system graphite2 and harfbuzz enabled).
Wed Oct 17 19:49:33 CDT 2018 harfbuzz-1.9.0
I have -1.8.8, but got the same problem when using the included harfbuzz.
The included version is also 1.8.8
To me, the only real candidate for the build issue is llvm.
Disagree - icu. My BLFS-8.2 has icu_60_2, which might be too old,
and I think that fedora disable system icu. I said "if I'm reading
it right" because that seemed an odd thing to do.
* = rebuilt at least once for printing support after CUPS was installed
NOTE - I do not have liboauth installed yet, and I don't have plans on
installing Doxygen or OpenJDK. I can if necessary though.
That magical combination of package versions lets me get to the end of
the build, up to the install process. I added --disable-elf-hack to my
mozconfig, and we'll see if that lets me get away with building it.
I tried --disable-elf-hack last night and it did not affect the error I was
getting.
I think Douglas's problem is different, but I don't know what is
causing it (unless he has a newer libelf than what is in the book).
I was looking at the error last night and it was looking for
void JSContext::checkImpl_63(...
Note the _63. From best I can tell, this is occurring from a generated
file, but the ./mach output makes it difficult to tell. I tried adding
--verbose, but did not seem to get a lot of info. I'll try Ken's
suggestion to use clang.
Indeed, _63 implies it comes from icu.
Without a doubt, there's a problem here somewhere.
Agreed.
-- Bruce
After the clang build (probably) fails, can you try commenting out
ac_add_options --with-system-icu
