On systems that don't have <stdint.h> (i.e., Microsoft, which is
tragically underfunded and cannot spare the resources necessary to
provide and support this header: http://tinyurl.com/absoh8),
SpiderMonkey header files should not introduce definitions for these
types, as doing so may conflict with client code's attempts to provide
its own definitions for these types.
Instead, have jstypes.h define JS{Int,Uint}{8,16,32,64,Ptr} types
based on configure's results, and make jsstdint.h into an uninstalled
header for use within SpiderMonkey that does whatever is necessary to
get definitions for the <stdint.h> types.
The changes to make the appropriate SpiderMonkey .cpp files #include
"jsstdint.h" explicitly are in a separate patch, for ease of review.
Delete jsstddef.h, since its only remaining purpose is to make certain
Win16-specific definitions; we don't support Win16 any more. In
particular, we can just subtract pointers now, so the PTRDIFF macro is
unnecessary noise.
Most places get stddef.h via jstypes.h or some other header, so we can
just delete #inclusions of jstddef.h. The exception is jskwgen.h, so
there we explicitly include <stddef.h> instead.
* Various code inside and outside of JS uses JS_BYTES_PER_WORD, so I added it to js-config.h
* Existing code uses JS_BYTES_PER_DOUBLE, JS_BITS_PER_WORD_LOG2, and JS_ALIGN_OF_POINTER so I've added autoconf tests for those
r=crowder r=jimb
* Various code inside and outside of JS uses JS_BYTES_PER_WORD, so I added it to js-config.h
* Existing code uses JS_BYTES_PER_DOUBLE, JS_BITS_PER_WORD_LOG2, and JS_ALIGN_OF_POINTER so I've added autoconf tests for those
r=crowder r=jimb
At configure time, check for <stdint.h>. If we don't have it, find
integer types of various sizes. On Windows, where we can't run
compilation tests in configure, hard-code definitions suggesting the
use of the built-in __intN types for the exact-size types, and
<stddef.h> for the pointer-sized types.
Use namespace-clean names for the preprocessor macros we define.
Since these types are used in the public JavaScript API, the configure
script needs to place the definitions it finds in js-config.h, the
installed configure-generated header, so it can be used by jsapi.h and
that gang.
New header js/src/jsstdint.h does what it takes to get definitions for
the exact-size and pointer-size integral types. It includes
<stdint.h> when available, uses the types found by configure.in to
define the {,u}int{8,16,32,64,ptr}_t types itself, or uses the __intN
types and the <stddef.h> header.
Remove now-unnecessary and possibly conflicting definitions of intN_t
types from js/src/nanojit/avmplus.h.
A directory's 'libs' target should create symlinks in the top-level
dist/bin and dist/lib directories to the libraries (and any other
generated files) in the build tree, so that when one has changed
sources in $src/js/src, simply running 'make' in $obj/js/src and then
re-running $obj/dist/bin/firefox will pick up the new libraries.
The js/src/Makefile.in 'install' target has many double-colon rules
with commands; we'd like to use these all in two circumstances:
. For 'make install' in a stand-alone js/src build. Since we're
following GNU behavior, this should copy the files it's installing.
. For 'make libs' in an in-tree js/src build. To interact properly
with the rest of the Mozilla build system, this should create
symlinks as explained above.
To accomplish this, we use plain old $(INSTALL) in all the install
targets, have js/src/Makefile.in set NSDISTMODE to copy (to get the
GNU behavior), and then have config/js/Makefile.in override that (to
get the Mozilla in-tree behavior).
Intel recommends against the use of -Os, and using it seems to produce
incorrect code in many recent versions of Intel's compilers.
js/src/Makefile.in tries to use -Os only with G++, but it tests
INTEL_CC, not INTEL_CXX --- even though almost all the sources are
C++. Check INTEL_CXX instead.
Record object files' dependency on javascript-trace.h, when
HAVE_DTRACE is set.
Use $(CURDIR) where javascript-trace.h appears as a target or
dependency, to ensure that VPATH doesn't cause us to use a copy other
than the one in the object directory.
SpiderMonkey now has its own copy of some of the files from ./config
and ./build. Since there is a decent amount of churn in that area, I
don't want it to become a burden to make merges back and forth. This
patch adds a comment explaining the 'identical if present' policy, and
runs a script to verify that it's actually being observed.
Have js/src/configure create a header file, js-config.h, that records
configure-controlled options that affect the SpiderMonkey API, like
'--enable-threadsafe'. js-config.h is namespace-clean, so it can be
installed with jsapi.h.
This means that clients can configure SpiderMonkey however they like,
and then simply #include "jsapi.h" and have everything work; they
don't have to remember to match their own compiler -D flags with those
SpiderMonkey's configure script chose. For example, mozilla-config.h
needn't concern itself with JS_THREADSAFE.
It seems to me this could also be done by having js-config --cflags
print -D options. The approach taken here seems a bit more robust: if
you can find jsapi.h at all, then you know you're getting the right
settings.
Give SpiderMonkey its own configure script and top-level Makefile.
Adjust js/src/Makefile as appropriate for life as a stand-alone
makefile, instead of a 'make export; make libs'-style Mozilla tier
makefile. Have the configure script accept '--with-nspr-cflags' and
'--with-nspr-libs' options for using an in-tree NSPR. Also accept
'--with-system-nspr', '--with-nspr-prefix', and
'--with-nspr-exec-prefix' flags for using an installed NSPR. Default
to --disable-jemalloc, assuming we don't have that part of the tree
available; have the top-level configure script pass --enable-jemalloc
as needed.
Since we no longer have an export phase to copy header files into
dist/include/js before we build the library, we need to be able to
find nanojit.h in the nanojit directory; fix references in
jsbuiltins.h and jstracer.cpp.
Give SpiderMonkey it its own copies of many of the files from ./config
and ./build. These are all exact copies, except as follows:
. js/src/config/Makefile.in: js/src only has a subset of
js/src/config, and thus a subset of the makefile targets.
. js/src/config/autoconf.mk.in: js/src/configure.in has its own make
variables to set, not set by the top-level configure script, so it
needs a custom automake.mk template.
. js/src/config/make-system-wrappers.pl: a copy from nsprpub/config,
so that we can build without having an NSPR source tree handy.
Invoke js/src/configure from ./configure, passing the values computed
for NSPR_CFLAGS and NSPR_LIBS by the top-level configure script.
Treat js/src as a static directory of the js tier, and create a new
config/js (just a Makefile) to be the js tier's non-static directory.
Let js/src/configure generate SpiderMonkey's makefiles, not
./configure.
Generate a 'js-config' script, which clients can call to find the
CFLAGS and LIBS values necessary to compile and link against an
installed SpiderMonkey library. Don't include the js-config script in
Macintosh packages.
Teach client.mk how to rebuild js/src/configure.
Tell Mercurial to ignore files generated by autoconf in js/src.
Further work:
. Right now, callers must define JS_THREADSAFE when #including jsapi.h.
This is fixed in a subsequent patch.
. js/src/configure is a trimmed copy of ./configure. It could be
trimmed more.
--HG--
rename : build/autoconf/acoutput-fast.pl => js/src/build/autoconf/acoutput-fast.pl
rename : build/autoconf/altoptions.m4 => js/src/build/autoconf/altoptions.m4
rename : build/autoconf/config.guess => js/src/build/autoconf/config.guess
rename : build/autoconf/config.sub => js/src/build/autoconf/config.sub
rename : build/autoconf/glib.m4 => js/src/build/autoconf/glib.m4
rename : build/autoconf/install-sh => js/src/build/autoconf/install-sh
rename : build/autoconf/make-makefile => js/src/build/autoconf/make-makefile
rename : build/autoconf/match-dir.sh => js/src/build/autoconf/match-dir.sh
rename : build/autoconf/nspr.m4 => js/src/build/autoconf/nspr.m4
rename : build/autoconf/pkg.m4 => js/src/build/autoconf/pkg.m4
rename : build/autoconf/update-makefile.sh => js/src/build/autoconf/update-makefile.sh
rename : build/cygwin-wrapper => js/src/build/cygwin-wrapper
rename : build/hcc => js/src/build/hcc
rename : build/hcpp => js/src/build/hcpp
rename : build/unix/mddepend.pl => js/src/build/unix/mddepend.pl
rename : build/unix/uniq.pl => js/src/build/unix/uniq.pl
rename : config/Makefile.in => js/src/config/Makefile.in
rename : config/Moz/Milestone.pm => js/src/config/Moz/Milestone.pm
rename : config/autoconf.mk.in => js/src/config/autoconf.mk.in
rename : config/config.mk => js/src/config/config.mk
rename : config/elf-dynstr-gc.c => js/src/config/elf-dynstr-gc.c
rename : config/fastcwd.pl => js/src/config/fastcwd.pl
rename : config/gcc_hidden.h => js/src/config/gcc_hidden.h
rename : config/insure.mk => js/src/config/insure.mk
rename : nsprpub/config/make-system-wrappers.pl => js/src/config/make-system-wrappers.pl
rename : config/milestone.pl => js/src/config/milestone.pl
rename : config/milestone.txt => js/src/config/milestone.txt
rename : config/mkdepend/Makefile.in => js/src/config/mkdepend/Makefile.in
rename : config/mkdepend/cppsetup.c => js/src/config/mkdepend/cppsetup.c
rename : config/mkdepend/def.h => js/src/config/mkdepend/def.h
rename : config/mkdepend/ifparser.c => js/src/config/mkdepend/ifparser.c
rename : config/mkdepend/ifparser.h => js/src/config/mkdepend/ifparser.h
rename : config/mkdepend/imakemdep.h => js/src/config/mkdepend/imakemdep.h
rename : config/mkdepend/include.c => js/src/config/mkdepend/include.c
rename : config/mkdepend/main.c => js/src/config/mkdepend/main.c
rename : config/mkdepend/mkdepend.man => js/src/config/mkdepend/mkdepend.man
rename : config/mkdepend/parse.c => js/src/config/mkdepend/parse.c
rename : config/mkdepend/pr.c => js/src/config/mkdepend/pr.c
rename : config/nfspwd.pl => js/src/config/nfspwd.pl
rename : config/nsinstall.c => js/src/config/nsinstall.c
rename : config/nsinstall.py => js/src/config/nsinstall.py
rename : config/pathsub.c => js/src/config/pathsub.c
rename : config/pathsub.h => js/src/config/pathsub.h
rename : config/preprocessor.pl => js/src/config/preprocessor.pl
rename : config/revdepth-nt.pl => js/src/config/revdepth-nt.pl
rename : config/revdepth.pl => js/src/config/revdepth.pl
rename : config/rules.mk => js/src/config/rules.mk
rename : config/system-headers => js/src/config/system-headers
rename : config/version.mk => js/src/config/version.mk
rename : config/version_win.pl => js/src/config/version_win.pl
rename : configure.in => js/src/configure.in
We'd like the SpiderMonkey configure script to generate a header file
named js-config.h to hold #definitions chosen at configure time that
affect the SpiderMonkey API, like JS_THREADSAFE. However, that name
is very similar to that of an existing header file, jsconfig.h. This
patch renames the existing header file, and updates all references to
it.
--HG--
rename : js/src/jsconfig.h => js/src/jsversion.h
from 2b107c27dedf (accidentally backed out during an earlier merge). 22 test
suite failures for -L lc2 lc3 spidermonkey-n slow-n, probably needs some guard
work in TRY_BRANCH_AFTER_COND.