* amd64-linux-nat.c: Include <sys/user.h>.
(ps_get_thread_area): Use PTRACE_PEEKUSER to get fs_base/gs_base
if HAVE_STRUCT_USER_REGS_STRUCT_FS_BASE or
HAVE_STRUCT_USER_REGS_STRUCT_GS_BASE is defined.
* configure.ac: Check if the fs_base and gs_base members of
`struct user_regs_struct' exist.
* config.in: Regenerated.
* configure: Likewise.
* configure.ac (development): Define new variable.
Call AC_CHECK_LIB for mcheck if $development.
(ERROR_ON_WARNING): Enable it by default only if $development.
* config.in: Regenerate.
* configure: Regenerate.
* NEWS (--with-auto-load-dir): Prepend $debugdir to the default path.
Describe it.
* auto-load.c (auto_load_expand_dir_vars): New function.
(auto_load_safe_path_vec_update): Use it, remove the
substitute_path_component call thanks to it.
(auto_load_objfile_script): Remove the debug_file_directory processing.
Use auto_load_expand_dir_vars, remove the substitute_path_component
call thanks to it.
* configure: Regenerate.
* configure.ac (--with-auto-load-dir): Prepend $debugdir to the default
path. Escape $ also for $debugdir.
(--with_auto_load_safe_path): Escape $ also for $debugdir.
* utils.c (substitute_path_component): Accept also DIRNAME_SEPARATOR.
gdb/doc/
* gdb.texinfo (Separate Debug Files): New anchor debug-file-directory.
Mention also --with-separate-debug-dir.
(Auto-loading): Prepend $debugdir in the sample output.
(Auto-loading safe path): Likewise. Mention also $debugdir for the
auto-load safe-path variable.
(objfile-gdb.py file): Remove the extra debug-file-directory paragraph.
Mention also $debugdir for 'set auto-load scripts-directory'.
Provide $ddir substitution for --with-auto-load-safe-path.
* NEWS (--with-auto-load-safe-path, --without-auto-load-safe-path): New
entries.
* auto-load.c: Include observer.h.
(auto_load_safe_path_vec_update): Call substitute_path_component for
each component. New variable ddir_subst.
(auto_load_gdb_datadir_changed): New function.
(set_auto_load_safe_path): Rename DEFAULT_AUTO_LOAD_SAFE_PATH to
AUTO_LOAD_SAFE_PATH. New comment.
(_initialize_auto_load): Rename DEFAULT_AUTO_LOAD_SAFE_PATH to
AUTO_LOAD_SAFE_PATH. Install auto_load_gdb_datadir_changed.
* config.in: Regenerate.
* configure: Regenerate.
* configure.ac (--auto-load-safe-path): Rename
DEFAULT_AUTO_LOAD_SAFE_PATH to AUTO_LOAD_SAFE_PATH. Default to
GDB_DATADIR/auto-load.
* defs.h (substitute_path_component): New declaration.
* top.c: Include observer.h.
(set_gdb_datadir): New function.
(init_main): Install it for "set data-directory".
* utils.c (substitute_path_component): New function.
gdb/doc/
Provide $ddir substitution for --with-auto-load-safe-path.
* gdb.texinfo (Auto-loading): Replace /usr/local by $ddir/auto-load.
(Auto-loading safe path): Likewise. Mention the default value,
$ddir substitution, --with-auto-load-safe-path and
--without-auto-load-safe-path.
* observer.texi (gdb_datadir_changed): New.
* auto-load.c (set_auto_load_safe_path): Reset AUTO_LOAD_SAFE_PATH
back to DEFAULT_AUTO_LOAD_SAFE_PATH if it is being set to "".
(show_auto_load_safe_path): Check any-directory by comparison with "/".
(add_auto_load_safe_path): Change the error message.
(_initialize_auto_load): Change the "safe-path" help text.
* configure: Regenerate
* configure.ac (--without-auto-load-safe-path): Set
WITH_AUTO_LOAD_SAFE_PATH to /.
gdb/doc/
* gdb.texinfo (Auto-loading safe path): Make 'directories'
for 'set auto-load safe-path' optional. Mention if it is omitted.
Change disabling security protection condition to "/", twice.
The problem showed up on AIX when we switched from the system linker
to GNU ld. What happens is that configure is adding -lpthdebug to
CONFIG_LDFLAGS when AIX thread debugging support is detected. But
this causes the "-lpthread" switch to be placed at the _before_
GDB's .o files (see Makefile.in):
INTERNAL_LDFLAGS = [...] $(CONFIG_LDFLAGS)
[...]
gdb$(EXEEXT): gdb.o $(LIBGDB_OBS) $(ADD_DEPS) $(CDEPS) $(TDEPLIBS)
rm -f gdb$(EXEEXT)
$(CC_LD) $(INTERNAL_LDFLAGS) $(WIN32LDAPP) \
-o gdb$(EXEEXT) gdb.o $(LIBGDB_OBS) \
$(TDEPLIBS) $(TUI_LIBRARY) $(CLIBS) $(LOADLIBES)
And as a result of this, the linker reports an error due to some
symbols in aix-thread.o not being satisfied.
This patch adds the "-lpthread" switch to LIBS instead, which should
be the right place to add it.
gdb/ChangeLog (Tristan Gingold):
* configure.ac (aix): Put -lpthread into libs.
* configure: Regenerate.
New option "set auto-load safe-path".
* NEWS: New commands "set auto-load safe-path"
and "show auto-load safe-path".
* auto-load.c: Include gdb_vecs.h, readline/tilde.h and completer.h.
(auto_load_safe_path, auto_load_safe_path_vec)
(auto_load_safe_path_vec_update, set_auto_load_safe_path)
(show_auto_load_safe_path, add_auto_load_safe_path, filename_is_in_dir)
(filename_is_in_auto_load_safe_path_vec, file_is_auto_load_safe): New.
(source_gdb_script_for_objfile): New variable is_safe. Call
file_is_auto_load_safe. Return if it is not.
(struct loaded_script): New field loaded.
(maybe_add_script): Add parameter loaded. Initialize SLOT with it.
(print_script): Use LOADED indicator instead of FULL_PATH. Change
output "Missing" to "No".
(_initialize_auto_load): New variable cmd. Initialize
auto_load_safe_path. Register "set auto-load safe-path",
"show auto-load safe-path" and "add-auto-load-safe-path".
* auto-load.h (maybe_add_script): Add parameter loaded.
(file_is_auto_load_safe): New declaration.
* config.in: Regenerate.
* configure: Regenerate.
* configure.ac: New parameters --with-auto-load-safe-path
and --without-auto-load-safe-path.
* linux-thread-db.c (try_thread_db_load_from_pdir_1)
(try_thread_db_load_from_dir): Check file_is_auto_load_safe first.
* main.c (captured_main): Check file_is_auto_load_safe for
LOCAL_GDBINIT.
* python/py-auto-load.c (gdbpy_load_auto_script_for_objfile): New
variable is_safe. Call file_is_auto_load_safe. Return if it is not.
(source_section_scripts): Call file_is_auto_load_safe. Return if it is
not.
gdb/doc/
New option "set auto-load safe-path".
* gdb.texinfo (Auto-loading): Extend the "show auto-load"
and "info auto-load" examples for safe-path. Put there also references
for "set auto-load safe-path" and "show auto-load safe-path".
New menu item for Auto-loading safe path.
(Auto-loading safe path): New node.
(Python Auto-loading): Update the expected output from "Missing"
to "No".
gdb/testsuite/
New option "set auto-load safe-path".
* gdb.python/py-objfile-script.exp (set auto-load safe-path): New.
* gdb.python/py-section-script.exp (set auto-load safe-path): New.
* libunwind-frame.c: Rename to ...
* ia64-libunwind-tdep.c: ... here.
* libunwind-frame.h: Rename to ...
* ia64-libunwind-tdep.h: ... here.
* Makefile.in (HFILES_NO_SRCDIR): Rename libunwind-frame.h to
ia64-libunwind-tdep.h.
(ALLDEPFILES): Rename libunwind-frame.c to ia64-libunwind-tdep.c.
* README (--with-libunwind): Rename to ...
(--with-libunwind-ia64): ... here, note it is ia64 specific now.
* config.in: Regenerate.
* configure: Regenerate.
* configure.ac: New option --with-libunwind-ia64, make the
AS_HELP_STRING ia64 specific. Deprecate option --with-libunwind.
Remove AC_DEFINE for HAVE_LIBUNWIND.
* ia64-libunwind-tdep.c: Make the file top comment ia64 specific.
Rename libunwind-frame.h #include to ia64-libunwind-tdep.h.
Rename libunwind-frame in the general comment.
* ia64-libunwind-tdep.h: Make the file top comment ia64 specific.
Rename symbol LIBUNWIND_FRAME_H to IA64_TDEP_LIBUNWIND_FRAME_H.
Move forward declarations inside #ifndef. Rename libunwind-frame in
the general comment.
* ia64-tdep.c: Rename libunwind-frame.h #include to
ia64-libunwind-tdep.h.
(ia64_gdb2uw_regnum, ia64_uw2gdb_regnum, ia64_is_fpreg)
(ia64_libunwind_descr): Rename libunwind-frame to
ia64-libunwind-tdep in these function comments.
* ia64-tdep.h: Rename libunwind-frame.h #include to
ia64-libunwind-tdep.h.
* ia64-vms-tdep.c (ia64_vms_libunwind_descr): Rename libunwind-frame to
ia64-libunwind-tdep in that data comment.
Pedro Alves <palves@redhat.com>
* ia64-tdep.c: Do not include libunwind-ia64.h.
* libunwind-frame.h: Remove #ifdef HAVE_LIBUNWIND_H guard.
Include libunwind-ia64.h instead of libunwind.h.
* configure.ac (--with-libunwind, $enable_libunwind): Don't check
for libunwind.h existence.
* configure, config.in: Regenerate.
* Makefile.in: Add gdb-dlfcn.c and gdb-dlfcn.h to build system.
* config.in: Add new #define HAVE_LIBDL.
* configure.ac: Add check for -ldl.
* configure: Re-generated by autoconf.
* gdb-dlfcn.c: New file.
* gdb-dlfcn.h: New file.
* Makefile.in: Add jit-reader.h as a header. Have it installed in
$(includedir)/gdb.
* configure.ac: Generate a correct value for TARGET_PTR for
jit-reader.h. Tell configure to generate jit-reader.h from
jit-reader.in.
* configure: Re-generated by autoconf.
* jit-reader.in: New file.
* jit.c: Include jit-reader.h.
Fix compatibility with texinfo versions older than 4.12.
* Makefile.in (MAKEINFO): Set to @MAKEINFO@.
(MAKEINFOFLAGS, MAKEINFO_EXTRA_FLAGS, MAKEINFO_CMD): New.
(MAKEHTMLFLAGS): Use MAKEINFO_CMD.
(FLAGS_TO_PASS): Add MAKEINFOFLAGS and MAKEINFO_EXTRA_FLAGS.
* configure: Regenerate.
* configure.ac (MAKEINFO): Find it, from libiberty/configure.ac.
(MAKEINFOFLAGS): Pre-set it to --split-size=5000000.
(MAKEINFO_EXTRA_FLAGS): New test for -DHAVE_MAKEINFO_CLICK.
gdb/doc/
Fix compatibility with texinfo versions older than 4.12.
* Makefile.in (MAKEINFO): Set to @MAKEINFO@.
(MAKEINFOFLAGS, MAKEINFO_EXTRA_FLAGS, MAKEINFO_CMD): New.
(MAKEHTMLFLAGS): Use MAKEINFO_CMD.
(gdb.info, gdbint.info, stabs.info, annotate.info): Use MAKEINFO_CMD.
* gdb.texinfo (Tail Call Frames): Convert @arrow{} to @click, when possible.
Make the conversion conditional on HAVE_MAKEINFO_CLICK, using variables
CALLSEQ1A, CALLSEQ1B, CALLSEQ2A and CALLSEQ2B.
On newer versions of AIX (6.x and later), this function is actually
declared in procinfo.h, thus causing a compilation warning when we
re-declare it ourselves. This patch adds a configure check for that
function allowing us to declare the function only if the declaration
isn't already present in one of procinfo system header.
gdb/ChangeLog:
PR gdb/12116:
* configure.ac: Add getthrds declaration check.
* configure, config.in: Regenerate.
* aix-thread.c (getthrds): Declare only if not already declared
in procinfo.h. More declaration out of get_signaled_thread to
global scope.
2011-03-09 Yao Qi <yao@codesourcery.com>
Revert:
2011-02-11 Yao Qi <yao@codesourcery.com>
* common/Makefile.in: Add copyright header.
2011-02-11 Yao Qi <yao@codesourcery.com>
* Makefile.in: Remove signals.o from COMMON_OBS. Link
libcommon.a.
* configure.ac: Add common to sub dir.
* configure: Regenerate.
gdb/gdbserver/
2011-03-09 Yao Qi <yao@codesourcery.com>
Revert:
2011-03-04 Yao Qi <yao@codesourcery.com>
* Makefile.in: Remove GNU make feature --directory.
2011-03-05 Yao Qi <yao@codesourcery.com>
* Makefile.in (CLEANDIRS, REQUIRED_SUBDIRS): New variable.
(subdir_do): New make target. Copied from gdb/Makefile.
(maintainer-clean, realclean, distclean, clean): Call corresponding
make targets in common/Makefile.
2011-02-11 Yao Qi <yao@codesourcery.com>
* configure.ac: Call AC_PROG_RANLIB.
* Makefile.in: Remove signals.o from OBS. Link libcommon.a.
* configure: Regenerate.
Non-GNU sed do not like the '?' quantifier when used in a s/// regexp
that involve back-references, causing the build to fail when trying
to link with Python support. This fixes it by using the '*' quantifier
instead.
gdb/ChangeLog:
* configure.ac: Work around non-GNU sed limitation when computing
python version number.
* configure: Regenerate.
This makes several adjustements to the configure python-config.py
scripts to deal with the differences between a Unix install of Python
and a Windows install of Python (as downloaded from the Python website).
Differences:
- The Python executable is directly in the python prefix directory
as opposed to inside the bin/ subdirectory.
- The name of the python library is does not have a dot in the version
number: On Unix, we have libpython2.7, while on Windows, it's
libpython27. So the regexp extracting the python version from
the Python lib filename had to be adjusted slightly.
Also, the tests checking the name of the libpython had to be
adjusted to allow for that.
- There are no link options following the -lpython<version> switch
on Windows, but the regexp extracting the python version was
using it as a delimiter. It had to be removed.
- python-config.py does not work on Windows, mostly because
some sysconfig variables are missing. They are not necessary
so the script was adapted to skip them if not defined.
- The paths returned by python-config.py follow the Windows filename
convention in terms of the directory separator, and this is causing
trouble when the build environment is cygwin (while the compiler
and Python are MinGW). We could have fixed that in the configure
script, but it felt simpler to do so in python-config.py
gdb/ChangeLog:
* configure.ac: Add handling of Python distribution on Windows.
* python-config.py: If the LIBS, SYSLIBS, LIBPL and/or LINKFORSHARED
sysconfig variables are not defined, then do not use them.
On Windows, if LIBPL is not defined, then use prefix + '/libs'
instead. On Windows, return all paths using forward-slashes
rather than backslashes.
This is preparation work for being able to build GDB with Python
support on MinGW.
So far, the "python<version>" subdirectory needs to be specified
when including a Python header file. In order to do that, we have
some special configury that tweaks the include path returned by
python-config.py such that the use of the subdirectory in the include
is necessary. This was done in order to protect ourselves from
possible filename conflicts, since some of the filenames chosen by
Python were a little generic.
The problem is that this cannot work with a standard Python install
on MinGW systems. On such systems, the .h files are located in
<python_prefix>/include. So, in preparation for allowing us to build
GDB on MinGW with Python support enabled, this patch changes the
requirement to provide the "python<version>" subdirectory in the
include directive.
The positive consequence is that we no longer need to have a set
of #include directives for each version of Python, since the include
directive is now the same for all versions of Python. However, the
downside is that we are losing the level of protection we were trying
to achieve by forcing the subdirectory in the include directive.
In order to reduce a bit the consequences of a possible conflict,
this patch also changes the location where the -I/path/to/python
switch goes, to be last in the list (suggested by Doug Evans).
One last change is the fact that we are now including Python.h
and all other Python include headers using angle brackets rather
than double-quotes. This fixes a problem on case-insensitive
systems where #include "Python.h" causes our gdb/python/python.h
header to be included instead of Python's <Python.h> header.
gdb/ChangeLog:
* configure.ac: Remove fallback behavior for building
against Python. Remove tweaking of Python include path.
Add PYTHON_CPPFLAGS and PYTHON_LIBS substitution.
(AC_TRY_LIBPYTHON): Adjust program used in linking test.
If link is successful, set PYTHON_CPPFLAGS and PYTHON_LIBS.
Always restore CPPFLAGS and LIBS after linking test.
* configure: Regenerated.
* Makefile.in (INTERNAL_CPPFLAGS): Add @PYTHON_CPPFLAGS@.
(INSTALLED_LIBS, CLIBS): Add @PYTHON_LIBS@.
* python/python-internal.h: Adjust includes of Python .h files.
This is a nasty interaction between Python and GDB. Basically,
Python causes some macros to be unilaterally defined in order
to turns some features on:
/* Define to activate features from IEEE Stds 1003.1-2001 */
#define _POSIX_C_SOURCE 200112L
/* Define to the level of X/Open that your system supports */
#define _XOPEN_SOURCE 600
But the problem is that they turn off defines provided by some
system headers on which we depend. Namely:
* sys/siginfo.h:
#if _SGIAPI
#define siginfo __siginfo
#endif
* sys/ucontext.h:
#if _SGIAPI && !defined(__SGI_NOUCONTEXT_COMPAT)
[...]
#define fp_r __fp_r
[...]
#define fp_csr __fp_csr
[...]
#endif
The important macro here is _SGIAPI, defined as follow in standards.h:
#define _SGIAPI ((defined(_SGI_SOURCE) && \
_NO_POSIX && _NO_XOPEN4 && _NO_XOPEN5) || \
(_ANSIMODE && _NO_POSIX && _NO_XOPEN4 && _NO_XOPEN5))
If one builds GDB without Python, then _SGIAPI is true, and all is fine.
But building with Python causes both _POSIX_C_SOURCE and _XOPEN_SOURCE
to trip all the _NO_[...] tests (_NO_POSIX, _NO_XOPEN4, _NO_XOPEN5).
And so we get build failures because we try to use undefined types, or
non-existent component names inside the regset structure.
The latter problem is observed only within irix5-nat.c, which means
that it is specific to IRIX. So it's easy to write the code in a way
that it does not require the macros (just use the real component names,
rather than relying on the macros to do the translation).
The former, on the other hand, is a little trickier, because the problem
occurs inside a generic unit (procfs.c). The solution I chose was to
adjust the configure script to add -Dsiginfo=__siginfo to the CPPFLAGS
if building with python using GCC on IRIX.
We hadn't seen this sort of issue up to now because the affect units
have not been dependent on the python includes up to now. Recent changes
have made them indirectly dependent on Python, thus triggering the issues.
gdb/ChangeLog:
* irix5-nat.c: Replace fp_r, fp_regs and fp_scr by __fp_r, __fp_regs
and __fp_scr respectively throughout.
* configure.ac: Compile with -Dsiginfo=__siginfo if building with
Python using GCC on IRIX.
* configure: Regenerate.
