This reverts commit bb2ab2647d.
It appears to break the use of `G_MODULE_BIND_LOCAL`, as evidenced by
the `test_local_binding()` test in `module-test.c` failing on FreeBSD
ever since that test was added. See #3415.
Looking back at the merge request (!3656), I should have pressed for
detail on why this change was needed.
Fixes: #3415
According to POSIX, the default behavior of `dlopen` is unspecified when
its flags include neither `RTLD_LOCAL` nor `RTLD_GLOBAL`. Consequently,
different platforms have different default behavior.
The default on Linux is `RTLD_LOCAL`, but the default on Darwin is
`RTLD_GLOBAL`. By passing `0` to `dlopen`, this results in the opposite
of the caller's intent when using `G_MODULE_BIND_LOCAL`.
Passing `RTLD_LOCAL` for `G_MODULE_BIND_LOCAL` allows the correct
behavior to be observed regardless of the platform's default.
G_MODULE_SUFFIX is deprecated now because you will get the wrong
results using it most of the time:
1. The suffix on macOS is usually 'dylib', but it's 'so' when using
Autotools, so there's no way to get the suffix correct using
a pre-processor macro.
2. Prefixes also vary in a platform-specific way. You may or may not have
a 'lib' prefix for the name on Windows and on Cygwin the prefix is
'cyg'.
3. The library name itself can vary per platform. For instance, you may
want to load foo-1.dll on Windows and libfoo.1.dylib on macOS. This
is for libraries, not modules, but that is still a use-case that
people use the GModule API for.
g_module_build_path() does take care of (2) on Cygwin, but it
fundamentally cannot handle the possibility of multiple options for
the module name, since it does not do any I/O. Hence, it is also
deprecated.
Instead, g_module_open() has been improved so that it takes care of
all this by searching the filesystem for combinations of possible
suffixes and prefixes on each platform. Along the way, the
documentation for it was also improved to make it clearer what it
does.
Closes https://gitlab.gnome.org/GNOME/glib/-/issues/520
Closes https://gitlab.gnome.org/GNOME/glib/-/issues/1413
These have all been added manually, as I’ve finished all the files which
I can automatically detect.
All the license headers in this commit are for LGPL-2.1-or-later, and
all have been double-checked against the license paragraph in the file
header.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #1415
Update several links to allow the remote to use its configured default
branch name, rather than specifying `master` as the default branch name.
This will help avoid breakage if any of these projects rename their
default branch in the future.
Fix a few of the links where they were hitting redirects or had moved.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Helps: #2348
Specifically, for uclibc at the moment. Other implementations may need
locking, but I haven’t checked any aside from uclibc-ng and glibc.
POSIX.1-2001 specifies that `dlerror()` is not thread-safe, but the
glibc (and likely other libc) implementation is.
Issue #399 was originally reported about eglibc, but that project has
since died and been merged back into glibc. So that’s one less thing to
worry about.
Signed-off-by: Philip Withnall <pwithnall@endlessos.org>
Fixes: #399
This was mostly machine generated with the following command:
```
codespell \
--builtin clear,rare,usage \
--skip './po/*' --skip './.git/*' --skip './NEWS*' \
--write-changes .
```
using the latest git version of `codespell` as per [these
instructions](https://github.com/codespell-project/codespell#user-content-updating).
Then I manually checked each change using `git add -p`, made a few
manual fixups and dropped a load of incorrect changes.
There are still some outdated or loaded terms used in GLib, mostly to do
with git branch terminology. They will need to be changed later as part
of a wider migration of git terminology.
If I’ve missed anything, please file an issue!
Signed-off-by: Philip Withnall <withnall@endlessm.com>
This is a partial change of the previous work[0].
On 64 bit Android since android-23, 'handle = dlopen(NULL); dlsym(handle)'
doesn't work. Instead, only 'dlsym(RTLD_DEFAULT)' returns a valid pointer.
However, RTLD_DEFAULT is defined as '(void *) 0x0' on 64bit Android which
is usually used for invalid value so this patch allows the specific case.
[0] 0d81bb4e31https://bugzilla.gnome.org/show_bug.cgi?id=788270
On 64 bit Android this is #defined to 0, which is considered an invalid
library handle in all other cases. RTLD_DEFAULT is only supposed to be
used with dlsym() it seems, and the usage here was just an
"optimization" before.
By dlopen'ing NULL, we get the same on all 64 bit Android variants and it
actually works instead of erroring out. On 32 bit Android, dlopen() of
NULL unfortunately usually gives us something useless that finds no
symbols whatsoever.
https://bugzilla.gnome.org/show_bug.cgi?id=776876
2007-11-07 Matthias Clasen <mclasen@redhat.com>
* *.c: Make header include order consistent (#71704,
Diego Escalante Urrelo)
svn path=/trunk/; revision=5817
Fri Feb 20 02:39:03 2004 Tim Janik <timj@gtk.org>
* applied patch from David Schleef <ds@schleef.org> which implements
a G_MODULE_BIND_LOCAL flag to g_module_open() to disable global
symbol registration.
Sat Mar 2 14:46:17 2002 Owen Taylor <otaylor@redhat.com>
* gmodule-beos.c (_g_module_build_path) * gmodule-dyld.c (_g_module_build_path)
gmodule-dl.c (_g_module_build_path): Use G_MODULE_SUFFIX rather than
a hardcoded ".so". (#72803)
Sat Feb 17 07:26:33 2001 Tim Janik <timj@gtk.org>
* configure.in (G_MODULE_HAVE_DLERROR): add check for broken RTLD_GLOBAL
(on OSF1 V5.0).
Sat Feb 17 07:27:15 2001 Tim Janik <timj@gtk.org>
* gmodule.c: work around platforms that have broken RTLD_GLOBAL.
2000-12-22 Sebastian Wilhelmi <wilhelmi@ira.uka.de>
* configure.in: Determine the suffix of the shared librarries for
this system. This is done analogous to
ltconfig.sh. G_MODULE_SUFFIX in glibconfig.h is set to either
"sl", "dll", or (most often) "so".
* tests/Makefile.am, tests/module-test.c,
tests/libmoduletestplugin_a.c, tests/libmoduletestplugin_b.c:
Added new testcase for gmodule. This is mostly copied from
gmodule/testgmodule.c, but unlike that is is quiet. (Why BTW are
some tests that verbose, not to say loquacious...)
* gmodule.c: Make g_module_open more tolerant wrt to the module
name. First it tries to open the module as named, if that fails,
it checks, whether it is a libtool archive and parses it, if that
fails it appends the systems shared library suffix
(i.e. ".so") (if not already found) and tries again and if that
fails it tries to append the ".la" libtool suffix (if not already
found) and parses it.
* gmodule.c: Lock recursive mutex during most module functions for
safety.
* gmodule-dl.c: Return an error from _g_module_symbol only, if
dlerror says so. All other functions return an error as well, if
dlerror returns NULL.
* testgmodule.c: Thanks to the above change the #ifdefs have
vanished.
* glib/glib-sections.txt: Added G_MODULE_SUFFIX.
* glib/tmpl/modules.sgml: Updated.
Wed Jul 26 12:59:31 2000 Tim Janik <timj@gtk.org>
* *.[hc]: applied patch from Andreas Persenius <ndap@swipnet.se> that
updates the license headers to the GNU Lesser General Public License,
as well as updating the copyright year to 2000.
Wed Mar 1 05:34:47 2000 Tim Janik <timj@gtk.org>
* gmodule-beos.c (_g_module_symbol): do not return NULL symbols.
* gmodule-os2.c: removed NetBSD specific defines.
(_g_module_self): set an error message for unsupported behaviour.
* gmodule-beos.c: many coding style fixups.
(_g_module_open):
(_g_module_self):
(_g_module_close):
(_g_module_symbol): bunch of memory leaks plugged.
* gmodule-dl.c: make sure the error message returned from dlerror()
is always != NULL, by using a wrapper function fetch_dlerror(). based
on a patch to fix _g_module_symbol() for NetBSD from Scott Presnell
<srp@zgi.com>.
* gmodule-dld.c: minor indentation.
* gmodule-win32.c: minor cleanups.
* merges from glib-1-2.
Wed Feb 24 00:08:42 CST 1999 Shawn T. Amundson <amundson@gtk.org>
* *.[ch]: inserted additional note to look for ChangeLog and
AUTHORS file for a log of modifications.
Thu Jan 21 12:40:11 EST 1999 Jeff Garzik <jgarzik@pobox.com>
* gmodule/gmodule-dl.c (_g_module_build_path):
Add braces to eliminate an ambiguous else warning.
* tests/{Makefile.am, string-test.c, strfunc-test.c}:
Separate string and strfunc tests, working towards goal of
having separate test for each of the GLib modules.
Add a couple GString length tests.
* README.win32: More text.
* config.h.win32 glibconfig.h.win32: Update to match the
corresponding generated files on Unix.
* makefile.msc: Update with new source files, and gthread
library. Use the compiler flag -MD instead of using -D_DLL and
"/nodefaultlib:libc msvcrt.lib" in the link phase.
* glib.def: Include new functions, drop removed ones.
* glib.h: Add comments about main loop and polling on Win32. (In
general, it's only for the GIMP's use.) Add Win32 IO Channel
functions. Remove the obsoleted old IO Channel stuff (which was
in #if 0 already).
* giowin32.c: New file.
* gmain.c: Include config.h, conditionalize <sys/time.h>
inclusion. Add g_poll implementation for Win32 (only for the
GIMP's needs for now, it's hard or even impossible to be as clean
and generic as on Unix). Implement g_get_current_time on Win32. If
threads aren't supported, don't try to wake up main thread's
loop. On Win32, use a semaphore and not a pipe to wake up the main
loop.
* gmessages.c: On Win32, allocate a console window if the standard
output handle is invalid before writing to stdout, and reopen stdout
to that console window.
* giochannel.c: Conditionalize unistd.h inclusion. Some indentation
cleanup.
* gstrfuncs.c: Include <signal.h>.
* gutils.c: On Win32, also check the HOMEDRIVE and HOMEPATH
environment variables.
* gmodule-dl.c gmodule-dld.c: In
_g_module_build_path, don't add the "lib" prefix and
".so" or ".sl" suffix if already there.
* gmodule-win32.c: Likewise for the ".dll" suffix.
* gthread-posix.c: Conditionalize <sys/time.h> inclusion.
branch. See the ChangeLog for details of the changes.
In brief overview:
- The set of threading functions can be set
- A default implementation is provided in -lgthread
- All static data structures are locked using these
functions if g_thread_init() is called.
Tue Oct 27 03:00:50 1998 Tim Janik <timj@gtk.org>
* glib.h: removed dummy structure definitions for struct _GCache,
_GTree, _GTimer, _GMemChunk, _GListAllocator and _GStringChunk.
* gutils.c: implement glib's inline functions _after_ all include
statements have been processed.
removed Tor's MAXPATHLEN check since there already was one supplied
further down in this file.
(LibMain): special cased the #ifdef __LCC__ case for NATIVE_WIN32,
since lcc maybe used on other platforms as well. why in hell is this
stuff required?
(g_get_any_init): for windows, if the user name is supplied, use it as
realname also.
in general, if there is no homedir specified, use the tmpdir that
we already figured.
* gtimer.c (g_timer_elapsed): changed a g_assert() statement to
g_return_if_fail().
* applied glib-tml-981020-0.patch for WIN32 portability, added some
comments and g_return_if_fail() statements, minor indentation fixes.
ChangeLog entry from Tor Lillqvist is appended.
* glib.h (struct dirent): use lower case structure members.
* glib.h:
* makefile.lcc:
* makefile.msc:
s/COMPILING_GLIB/GLIB_COMPILATION/
1998-10-20: Tor Lillqvist <tml@iki.fi>
* README.win32 glib.def gmodule.def
* glibconfig.h.win32 gmodule/gmoduleconf.h.win32:
New files for the Windows port. The .def files list exported
symbols for the Microsoft linker and compatibles.
* configure.in:
Added checks for some platform-dependent headers: pwd.h sys/param.h
sys/select.h sys/time.h sys/times.h unistd.h, and the function lstat.
* gerror.c:
Conditionalized inclusion of system-dependent headers. Changes
for Windows: no gdb to do a stack trace. Just call abort().
* glib.h:
Changes for Windows:
Added macros G_DIR_SEPARATOR, G_DIR_SEPARATOR_S for
platform-dependent file name syntax elements. Added macros
G_SEARCHPATH_SEPARATOR, G_SEARCHPATH_SEPARATOR_S for
platform-dependent search path syntax conventions.
Added pragmas for Microsoft C to make it more pedantic.
Marked GLib's global variables for export from DLL.
Added the function g_strescape that escapes backslashes.
Added functions g_path_is_absolute and g_path_skip_root to
handle platform-dependent file name syntax.
Added the function g_getenv that expands environment variables
that contain references to other environment variables, as is
typical on Windows NT.
Added the GIOChannel structure which is used to encapsulate the
IPC mechanism used by the GIMP's plug-ins, and possibly other
things later. On Unix a GIOChannel encapsulates just a file
descriptor. On Windows it contains a file handle from _pipe() and a
few other things related to the implementation of gdk_input_add
and GIMP plug-in communication. Subject to change.
Removed duplicate declarations of the version variables.
For the Microsoft compiler, declare own implementation of
ftruncate and the <dirent.h> functions.
* gmem.c:
Define a symbolic name for the profiling table size.
* gmessages.c:
Conditionalized inclusion of unistd.h. On Windows, output using
stdio to stdout.
* gscanner.c:
Conditionalized inclusion of unistd.h. Added changes for
Microsoft C. Added CR to the skipped character set. Added small
workaround for MSC compiler bug in g_scanner_cur_value.
* gstrfuncs.c:
Added the function g_strescape, which escapes the backslash
character. Needed especially when printing Windows filenames.
* gtimer.c:
Conditionalized inclusion of unistd.h and sys/time.h. Added
implementations for Windows.
* gutils.c:
Conditionalized inclusion of platform-dependent headers. Use
the platform-independent file name syntax macros.
Conditionalize code on platform-dependent features. Added the
functions g_path_is_absolute g_path_skip_root and g_getenv.
Added the GIOChannel-related functions. Added
compiler-dependent Unix compatibility functions for Windows.
* makefile.lcc makefile.msc:
New files. Compiler-specific makefiles for LCC-Win32 and
Microsoft C. Only Microsoft C is actually supported currently.
* testglib.c:
Added pathname check cases for Windows. Added workaround for
bug in the Microsoft runtime library. Improved some tests a bit.
Tue Oct 27 04:00:11 1998 Tim Janik <timj@gtk.org>
* testgmodule.c (main): changed the #ifdef WIN32 test to NATIVE_WIN32,
this needs to be more constistent throughout the code, do we go for
NATIVE_WIN32 or WIN32?
* gmodule.c (LibMain): special cased the #ifdef __LCC__ case for
NATIVE_WIN32, since lcc maybe used on other platforms as well.
* libgplugin_a.c (LibMain):
* libgplugin_b.c (LibMain):
likewise. not sure i like this special requirement for lcc in here.
* gmodule-dl.c (_g_module_build_path):
feature empty "" directories and prepend the module name with "lib".
* gmodule-dld.c (_g_module_build_path):
* gmodule-win32.c (_g_module_build_path):
feature empty "" directories.
* we need some more magic in the _g_module_build_path variants
so we don't append/prepend lib and .so, .sl or .dll for those names
that already contain it.
* applied patch from Tor Lillqvist for g_module_build_path() and
windows support.
1998-10-20: Tor Lillqvist <tml@iki.fi>
* gmodule/gmodule-win32.c:
New file.
* gmodule/gmodule.c gmodule/gmodule.h:
Added the funcion g_module_build_path that builds the path to
a module file, decorating the name according to the system's
conventions. Added the Windows implementation.
* gmodule/libgplugin_a.c gmodule/libgplugin_b.c:
Added LibMain for LCC-Win32.
* gmodule/testgmodule.c:
Handle Windows dll names.
Mon Aug 10 03:35:57 1998 Tim Janik <timj@gtk.org>
* gmodule.c: minor changes to internal interface.
* gmodule-dl.c:
* gmodule-dld.c: put some comments into the files, and provided
better error checking for shl_findsym(). whish i had a system to
test this stuff on.
