When trying to build gdbserver on ppc-lynx178, the compiler reports
while trying to compile gdbserver/ax.c that vsprintf is not declared.
Looking at my C99 reference manual (a draft), I see the following
synopsis:
#include <stdarg.h>
#include <stdio.h>
int vsprintf(char * restrict s, [etc]);
Looking at stdio.h on LynxOS-178, if found where vsprintf gets
declared:
#if defined(__varargs_h) || defined(__stdarg_h) \
|| defined(_VARARGS_H) || defined(_STDARG_H)
extern int vsprintf _AP((char *, const char *, va_list));
#endif
Digging further, I noticed that common-defs.h, which is included
via server.h, includes stdarg.h after including stdio, explaining
why vsprintf does not get declared in this case.
This patch fixes the problem by including stdarg.h before stdio.h.
gdb/ChangeLog:
* common/common-defs.h: Move <stdarg.h> #include ahead of
<stdio.h> #include.
Tested on x86_64-linux.
We're currently pulling gnulib's errno module as a dependency of some
other module. That provides an errno.h that defines EILSEQ to a
distinct value if the system's errno.h doesn't define it already.
However, GNU iconv does this:
/* Get errno declaration and values. */
#include <errno.h>
/* Some systems, like SunOS 4, don't have EILSEQ. Some systems, like BSD/OS,
have EILSEQ in a different header. On these systems, define EILSEQ
ourselves. */
#ifndef EILSEQ
#define EILSEQ @EILSEQ@
#endif
That's in:
http://git.savannah.gnu.org/cgit/libiconv.git/tree/include/iconv.h.in
The "different header" mentioned is wchar.h. This is handled in:
http://git.savannah.gnu.org/cgit/libiconv.git/tree/m4/eilseq.m4
which defines @EILSEQ@ to ENOENT if EILSEQ isn't found in either
errno.h or wchar.h.
So if iconv sets errno to EILSEQ on such system's, it's really setting
it to ENOENT. And when we check for EILSEQ, we're checking for
gnulib's value. The result is we won't detect the error correctly.
As we dropped support for both SunOS 4 or old BSD/OS, maybe we don't
need to care about the wchar.h issue anymore. Still, AFAICS, gnulib's
m4/errno_h.m4 doesn't know that EILSEQ may be defined in wchar.h, and
so on such systems, ISTM gnulib ends up defining an incompatible
EILSEQ itself, but I think that should be fixed on the gnulib side, by
making it extract the EILSEQ value out of the system's wchar.h, like
GNU iconv does.
So that leaves handling the case of gnulib making up a EILSEQ value,
which we take as meaning the system really doesn't really define it,
which will be the same systems GNU iconv sets errno to ENOENT instead
of EILSEQ.
Looking at glibc's iconv it seems that ENOENT is never used there.
It seems it's safe to always treat ENOENT the same as EILSEQ.
The current EILSEQ definition under PHONY_ICONV is obviously stale as
gnulib garantees there's always a EILSEQ defined.
Tested on x86_64 Fedora 20.
gdb/
2014-11-14 Pedro Alves <palves@redhat.com>
* charset.c [PHONY_ICONV && !EILSEQ] (EILSEQ): Don't define.
[!PHONY_ICONV] (gdb_iconv): New function.
[!PHONY_ICONV] (iconv): Redefine to gdb_iconv.
PR binutils/17512
* dwarf.c (get_encoded_value): Add an 'end' parameter. Change the
'data' parameter to a double pointer and return the updated value.
(decode_location_expression): Update call to get_encoded_value.
(frame_need_space): Handle the case where one or both of the
mallocs fails.
(read_cie): Initialise the cie pointer, even if the read fails.
(display_debug_frames): Warn if the calculated block_end is before
the start of the block. Break the loop if the CIE could not be
read. Update call to get_encoded_value. Warn if the read CFA
expressions are too big.
* ieee.c (ieee_archive_p) Skip processing if no bytes are read at
all.
(ieee_object_p): Likewise.
The patch <https://sourceware.org/ml/gdb-patches/2014-03/msg00202.html>
fixed dw2-ifort-parameter.exp on powerpc64 by adding some labels to
get the start and end address of function func. This should also fix the
fail on thumb mode, however, this style is quite specific to gcc, and
other compiler, such as clang, may not guarantee the order of global
asms and functions. The test fails with clang:
$ make check RUNTESTFLAGS='dw2-ifort-parameter.exp CC_FOR_TARGET=clang'
(gdb) p/x param^M
No symbol "param" in current context.^M
(gdb) FAIL: gdb.dwarf2/dw2-ifort-parameter.exp: p/x param
With this patch applied, dw2-ifort-parameter.exp still passes for gcc
on arm thumb mode and popwerpc64, and it also passes for clang on
x86_linux.
gdb/testsuite:
2014-11-14 Yao Qi <yao@codesourcery.com>
* gdb.dwarf2/dw2-ifort-parameter.c: Remove inline asm.
(func): Add label func_label.
* gdb.dwarf2/dw2-ifort-parameter.exp (Dwarf::assemble):
Replace low_pc and high_pc with MACRO_AT_range.
Replace name, low_pc and high_pc with MACRO_AT_func.
Hi,
I see the fail in gdb.dwarf2/implptr-optimized-out.exp in thumb mode
(gdb) p p->f^M
No symbol "p" in current context.^M
(gdb) FAIL: gdb.dwarf2/implptr-optimized-out.exp: p p->f
and the crash on powerpc64
(gdb) continue^M
Continuing.^M
^M
Program received signal SIGSEGV, Segmentation fault.^M
0x7d82100810000828 in ?? ()
The cause of both is that we incorrectly set attribute low_pc, since
main isn't resolved to function start address on these targets.
In this patch, we replace attributes name, low_pc and high_pc with
MACRO_AT_func. The fail on thumb mode is fixed, and crash on
powerpc64 is fixed too.
gdb/testsuite:
2014-11-14 Yao Qi <yao@codesourcery.com>
* gdb.dwarf2/implptr-optimized-out.exp (Dwarf::assemble):
Replace name, low_pc and high_pc with MACRO_AT_func.
This patch is to use dwarf::assemble to generate debug information, and
remove implptr-optimized-out.S as a result.
gdb/testsuite:
2014-11-14 Yao Qi <yao@codesourcery.com>
* gdb.dwarf2/implptr-optimized-out.exp: Use Dwarf::assemble to
produce debug information.
* gdb.dwarf2/implptr-optimized-out.S: Removed.
On arm-none-eabi target thumb mode, I see the following fail,
p the_int^M
$2 = 99^M
(gdb) FAIL: gdb.dwarf2/dwz.exp: p the_int
and on powerpc64 target, we even can't get function main from object
file,
disassemble main^M
No function contains specified address.^M
(gdb) FAIL: gdb.dwarf2/dwz.exp: disassemble main
This patch is to use MACRO_AT_func attribute to get the main's start
address and end address correctly, and also remove some code dwz.exp
getting main's length. This patch fixes fails on both thumb mode and
powerpc64 target.
PASS: gdb.dwarf2/dwz.exp: p other_int
PASS: gdb.dwarf2/dwz.exp: p the_int
gdb/testsuite:
2014-11-14 Yao Qi <yao@codesourcery.com>
* gdb.dwarf2/dwz.exp: Remove the code to compile main.c to
object and get function length.
(Dwarf::assemble): Replace name, low_pc and high_pc attributes
with MACRO_AT_func.
(top-level): Replace gdb_compile and clean_restart with
prepare_for_testing.
* gdb.dwarf2/main.c (main): Add label main_label.
This patch addes DW macro attributes MACRO_AT_func and MACRO_AT_range
in dwarf assembler, which emits "DW_AT_low_pc func_start addr" and
"DW_AT_high_pc func_end addr". func_start and func_end are computed
automatically by proc function_range.
These two attributes are pseudo attribute or macro attribute, which
means they are not standard dwarf attribute in dwarf spec. Then can
be substituted or expanded to standard attributes or macro attributes.
See details in the comments to them. Dwarf assembler is extended to
handle them.
Now the attributes name/low_pc/high_pc can be replaced with
MACRO_AT_func like this:
subprogram {
{name main}
{low_pc main_start addr}
{high_pc main_end addr}
}
becomes:
subprogram {
{MACRO_AT_func { main ${srcdir}/${subdir}/${srcfile} }}
}
users don't have to worry about the start and end of function main, and
they only need to add a label main_label in main.
gdb/testsuite:
2014-11-14 Yao Qi <yao@codesourcery.com>
* lib/dwarf.exp (function_range): New procedure.
(Dwarf::_handle_macro_at_func): New procedure.
(Dwarf::_handle_macro_at_range): New procedure.
(Dwarf): Handle MACRO_AT_func and MACRO_AT_range.
This patch is to move some code to a new procedure _handle_attribute,
which will be used in my following patches.
gdb/testsuite:
2014-11-14 Yao Qi <yao@codesourcery.com>
* lib/dwarf.exp (_handle_DW_TAG): Move some code to ...
(_handle_attribute): New procedure.
PR binutils/17531
* readelf.c (process_version_sections): If the read of the version
def information fails, make sure that the external verdef data is
not used.
(get_dynamic_data): Do not attempt to allocate memory for more
dynamic data than there is in the file. If the read fails, free
the allocated buffer.
(process_symbol_table): Do not print dynamic information if we
were unable to read the dynamic symbol table.
(print_gnu_note): Do not print the note if the descsz is too
small.
Remove old-style function header variants from sepdebug.c. Eliminate
references to the removed locations "breakpoint 9" and "breakpoint 13"
from sepdebug.exp.
gdb/testsuite/ChangeLog:
* gdb.base/sepdebug.c: Remove #ifdef PROTOTYPES, keep prototyped
variant.
* gdb.base/sepdebug.exp: Drop references to removed code.
Remove old-style function header variants from list0.h and list1.c.
Fill the removed lines with comments or empty lines, such that the
line numbering is undisturbed. Changes to the line numbering would
require heavy adjustments to list.exp, where many line numbers are
hard-coded, as well as a fair amount of knowledge about the source
code in and around certain lines. Thus the dependency on the line
numbering can not be eliminated so easily, and it may not even be a
useful goal for a "list" test case. Another option might be to adjust
the literal line numbers in list.exp, but even that is not as
straightforward as it may seem, since the test case expects certain
source lines to be exactly n lines apart.
gdb/testsuite/ChangeLog:
* gdb.base/list0.h: Remove #ifdef PROTOTYPES, keep prototyped
variant. Preserve original line numbering.
* gdb.base/list1.c: Likewise.
Remove old-style function headers from break.c and break1.c. Adjust
break.exp accordingly; in particular eliminate references to the
removed locations "breakpoint 9, 13, and 16" from break.exp.
gdb/testsuite/ChangeLog:
* gdb.base/break.c: Remove #ifdef PROTOTYPES, keep prototyped
variant.
* gdb.base/break1.c: Likewise.
* gdb.base/break.exp: Drop references to removed code.
The previous patch did not indent perform_all_tests() correctly after
moving the main logic into it, to avoid obscuring the functional
changes. This patch fixes the indentation.
gdb/testsuite/ChangeLog:
* gdb.base/callfuncs.exp (perform_all_tests): Re-indent.
In callfuncs.exp, compile callfuncs.c with and without C function
header prototypes and execute all tests after each compilation.
gdb/testsuite/ChangeLog:
* gdb.base/callfuncs.exp: Remove 'prototypes' variable. Move main
logic into perform_all_tests() and invoke it with and without
function header prototypes.
(do_function_calls): Remove conditional XFAIL for PR 5318.
(rerun_and_prepare): Remove duplicate code.
(perform_all_tests): New. Main logic moved here.
The C source file for the 'callfuncs' test case did not compile with
-DNO_PROTOTYPES or -DPROTOTYPES. This patch fixes various syntax
errors under #ifdef NO_PROTOTYPES and a small typo under #ifdef
PROTOTYPES.
gdb/testsuite/ChangeLog:
* gdb.base/callfuncs.c (t_float_many_args): Fix syntax error in
code guarded by #ifdef NO_PROTOTYPES.
(t_double_many_args): Likewise.
(DEF_FUNC_MANY_ARGS_1): Likewise.
(DEF_FUNC_VALUES_1): Likewise.
(t_structs_ldc): Renamed from t_structs_fc in conditional code
guarded by #ifdef PROTOTYPES.
Remove the literal line number from a regexp in mi-console.exp. Add
an appropriate eye-catcher to mi-console.c and refer to that instead.
gdb/testsuite/ChangeLog:
* gdb.mi/mi-console.c: Add eye-catcher.
* gdb.mi/mi-console.exp (semihosted_string): Refer to eye-catcher
instead of literal line number.
Remove the literal line number from a regexp in shlib-call.exp. Add
an appropriate eye-catcher to shr2.c and refer to that instead.
gdb/testsuite/ChangeLog:
* gdb.base/shr2.c: Add eye-catcher.
* gdb.base/shlib-call.exp: Refer to eye-catcher instead of literal
line number.
Remove literal line numbers from the regexps in jump.exp. Add
appropriate eye-catchers to jump.c and refer to those instead.
gdb/testsuite/ChangeLog:
* gdb.base/jump.c: Add eye-catchers.
* gdb.base/jump.exp: Refer to eye-catchers instead of literal line
numbers.
Remove literal line numbers from the regexps in foll-exec.exp. Add
appropriate eye-catchers to foll-exec.c and execd-proc.c and refer to
those instead.
gdb/testsuite/ChangeLog:
* gdb.base/execd-prog.c: Add eye-catchers.
* gdb.base/foll-exec.c: Likewise.
* gdb.base/foll-exec.exp: Refer to eye-catchers instead of literal
line numbers.
Remove literal line numbers from the regexps in ending-run.exp. Add
appropriate eye-catchers to ending-run.c and refer to those instead.
gdb/testsuite/ChangeLog:
* gdb.base/ending-run.c: Add eye-catchers.
* gdb.base/ending-run.exp: Refer to eye-catchers instead of
literal line numbers.
Remove literal line numbers from the regexps in call-rt-st.exp. Add
appropriate eye-catchers to call-rt-st.c and refer to those instead.
gdb/testsuite/ChangeLog:
* gdb.base/call-rt-st.c: Add eye-catchers.
* gdb.base/call-rt-st.exp: Refer to eye-catchers instead of
literal line numbers.
Remove literal line numbers from the regexps in call-ar-st.exp. Add
appropriate eye-catchers to call-ar-st.c and refer to those instead.
gdb/testsuite/ChangeLog:
* gdb.base/call-ar-st.c: Add eye-catchers.
* gdb.base/call-ar-st.exp: Refer to eye-catchers instead of
literal line numbers.
Remove literal line numbers from the commands and regexps in dbx.exp.
Add appropriate eye-catchers to average.c and sum.c and refer to those
instead.
gdb/testsuite/ChangeLog:
* gdb.base/average.c: Add eye-catchers.
* gdb.base/sum.c: Likewise.
* gdb.base/dbx.exp: Use eye-catchers to determine line numbers for
regexps dynamically.
Remove literal line numbers from the regexps in so-impl-ld.exp. Add
appropriate eye-catchers to solib1.c and refer to those instead.
gdb/testsuite/ChangeLog:
* gdb.base/solib1.c: Add eye-catchers.
* gdb.base/so-impl-ld.exp: Match against eye-catchers instead of
literal line numbers.
PR binutils/17512
* dwarf.c (read_and_display_attr_value): Check that we do not read
past end.
(display_debug_pubnames_worker): Add range checks.
(process_debug_info): Check for invalid pointer sizes.
(display_loc_list): Likewise.
(display_loc_list_dwo): Likewise.
(display_debug_ranges): Likewise.
(display_debug_aranges): Check for invalid address size.
(read_cie): Add range checks. Replace call strchr with while loop.
* objdump.c (dump_dwarf): Replace abort with a warning message.
(print_section_stabs): Improve range checks.
* rdcoff.c (coff_get_slot): Use long for indx parameter type.
Add check for an excesively large index.
* rddbg.c (read_section_stabs_debugging_info): Zero terminate the
string table. Avoid walking off the end of the stabs data.
* stabs.c (parse_stab_string): Add check for a NULL name.
PR binutils/17512
* coffcode.h (coff_slurp_line_table): Set the line number of
corrupt entries to -1.
(coff_slurp_symbol_table): Alway initialise the value of the
symbol.
* coffgen.c (coff_print_symbol): Check that the combined pointer
is valid.
(coff_print_symbol): Do not print negative line numbers.
* peXXigen.c (pe_print_idata): Add range checking displaying
member names.
As no place in the backends check cont_thread anymore, we can stop
setting and clearing it in places that resume the target and wait for
events. Instead simply clear it whenever a new GDB connects.
gdb/gdbserver/
2014-11-12 Pedro Alves <palves@redhat.com>
* server.c (cont_thread): Update comment.
(start_inferior, attach_inferior): No longer clear cont_thread.
(handle_v_cont): No longer set cont_thread.
(captured_main): Clear cont_thread each time a GDB connects.
There's code in linux_wait_1 that resumes all threads if the Hc thread
disappears. It's the wrong thing to do, as GDB has told GDBserver to
resume only one thread, because e.g., the user has scheduler-locking
enabled, or because GDB was stepping the program over a breakpoint.
Resuming all threads behind GDB's back can't be good in either case.
The right thing to do is to detect that that the (only) resumed thread
is gone, and let GDB know about it. The Linux backend is already
doing that nowadays, since:
commit fa96cb382c
Author: Pedro Alves <palves@redhat.com>
AuthorDate: Thu Feb 27 14:30:08 2014 +0000
Teach GDBserver's Linux backend about no unwaited-for children (TARGET_WAITKIND_NO_RESUMED).
The backend detects that all resumed threads have disappeared, and
returns TARGET_WAITKIND_NO_RESUMED to the core of GDBserver, which
then reports an error to GDB.
There's no need to frob the passed in ptid to wait for the continue
thread either -- linux_wait_for_event only returns events for resumed
threads.
The badness (of resuming threads) can actually be observed in the
testsuite, if we force-disable vCont support in GDBserver -- before
the patch, gdb.threads/no-unwaited-for-left.exp hangs if we disable
vCont:
(gdb) continue
Continuing.
FAIL: gdb.threads/no-unwaited-for-left.exp: continue to breakpoint: break-here (timeout)
... more cascading timeouts ....
After the patch, gdb.threads/no-unwaited-for-left.exp behaves the same
with or without vCont support:
(gdb) continue
Continuing.
[New Thread 32226]
[Switching to Thread 32226]
Breakpoint 2, thread_a (arg=0x0) at /home/pedro/gdb/mygit/build/../src/gdb/testsuite/gdb.threads/no-unwaited-for-left.c:28
28 return 0; /* break-here */
(gdb) PASS: gdb.threads/no-unwaited-for-left.exp: continue to breakpoint: break-here
...
continue
Continuing.
warning: Remote failure reply: E.No unwaited-for children left.
[Thread 32222] #1 stopped.
(gdb) FAIL: gdb.threads/no-unwaited-for-left.exp: continue stops when the main thread exits
Overall, this is also good for getting rid of a RSP detail from the backend.
gdb/gdbserver/
2014-11-12 Pedro Alves <palves@redhat.com>
* linux-low.c (linux_wait_1): Don't force a wait for the Hc
thread, and don't resume all threads if the Hc thread has exited.
The target->request_interrupt callback implements the handling for
ctrl-c. User types ctrl-c in GDB, GDB sends a \003 to the remote
target, and the remote targets stops the program with a SIGINT, just
like if the user typed ctrl-c in GDBserver's terminal.
The trouble is that using kill_lwp(signal_pid, SIGINT) sends the
SIGINT directly to the program's main thread. If that thread has
exited already, then that kill won't do anything.
Instead, send the SIGINT to the process group, just like GDB
does (see inf-ptrace.c:inf_ptrace_stop).
gdb.threads/leader-exit.exp is extended to cover the scenario. It
fails against GDBserver before the patch.
Tested on x86_64 Fedora 20, native and GDBserver.
gdb/gdbserver/
2014-11-12 Pedro Alves <palves@redhat.com>
* linux-low.c (linux_request_interrupt): Always send a SIGINT to
the process group instead of to a specific LWP.
gdb/testsuite/
2014-11-12 Pedro Alves <palves@redhat.com>
* gdb.threads/leader-exit.exp: Test sending ctrl-c works after the
leader has exited.
No longer used since the non-continuable watchpoints handling rework.
gdb/
2014-11-12 Pedro Alves <palves@redhat.com>
* infrun.c (enum infwait_states, infwait_state): Delete.
The gdb.arch/i386-bp_permanent.exp test is currently failing an
assertion recently added:
(gdb) stepi
../../src/gdb/infrun.c:2237: internal-error: resume: Assertion `sig != GDB_SIGNAL_0' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n)
FAIL: gdb.arch/i386-bp_permanent.exp: Single stepping past permanent breakpoint. (GDB internal error)
The assertion expects that the only reason we currently need to step a
breakpoint instruction is when we have a signal to deliver. But when
stepping a permanent breakpoint (with or without a signal) we also
reach this code.
The assertion is correct and the permanent breakpoints skipping code
is wrong.
Consider the case of the user doing "step/stepi" when stopped at a
permanent breakpoint. GDB's `resume' calls the
gdbarch_skip_permanent_breakpoint hook and then happily continues
stepping:
/* Normally, by the time we reach `resume', the breakpoints are either
removed or inserted, as appropriate. The exception is if we're sitting
at a permanent breakpoint; we need to step over it, but permanent
breakpoints can't be removed. So we have to test for it here. */
if (breakpoint_here_p (aspace, pc) == permanent_breakpoint_here)
{
gdbarch_skip_permanent_breakpoint (gdbarch, regcache);
}
But since gdbarch_skip_permanent_breakpoint already advanced the PC
manually, this ends up executing the instruction that is _after_ the
breakpoint instruction. The user-visible result is that a single-step
steps two instructions.
The gdb.arch/i386-bp_permanent.exp test is actually ensuring that
that's indeed how things work. It runs to an int3 instruction, does
"stepi", and checks that "leave" was executed with that "stepi". Like
this:
(gdb) b *0x0804848c
Breakpoint 2 at 0x804848c
(gdb) c
Continuing.
Breakpoint 2, 0x0804848c in standard ()
(gdb) disassemble
Dump of assembler code for function standard:
0x08048488 <+0>: push %ebp
0x08048489 <+1>: mov %esp,%ebp
0x0804848b <+3>: push %edi
=> 0x0804848c <+4>: int3
0x0804848d <+5>: leave
0x0804848e <+6>: ret
0x0804848f <+7>: nop
(gdb) si
0x0804848e in standard ()
(gdb) disassemble
Dump of assembler code for function standard:
0x08048488 <+0>: push %ebp
0x08048489 <+1>: mov %esp,%ebp
0x0804848b <+3>: push %edi
0x0804848c <+4>: int3
0x0804848d <+5>: leave
=> 0x0804848e <+6>: ret
0x0804848f <+7>: nop
End of assembler dump.
(gdb)
One would instead expect that a stepi at 0x0804848c stops at
0x0804848d, _before_ the "leave" is executed. This commit changes GDB
this way. Care is taken to make stepping into a signal handler when
the step starts at a permanent breakpoint instruction work correctly.
The patch adjusts gdb.arch/i386-bp_permanent.exp in this direction,
and also makes it work on x86_64 (currently it only works on i*86).
The patch also adds a new gdb.base/bp-permanent.exp test that
exercises many different code paths related to stepping permanent
breakpoints, including the stepping with signals cases. The test uses
"hack/trick" to make it work on all (or most) platforms -- it doesn't
really hard code a breakpoint instruction.
Tested on x86_64 Fedora 20, native and gdbserver.
gdb/
2014-11-12 Pedro Alves <palves@redhat.com>
* infrun.c (resume): Clear the thread's 'stepped_breakpoint' flag.
Rewrite stepping over a permanent breakpoint.
(thread_still_needs_step_over, proceed): Don't set
stepping_over_breakpoint for permanent breakpoints.
(handle_signal_stop): Don't clear stepped_breakpoint. Also pull
single-step breakpoints out of the target on hardware step
targets.
(process_event_stop_test): If stepping a permanent breakpoint
doesn't hit the step-resume breakpoint, delete the step-resume
breakpoint.
(switch_back_to_stepped_thread): Also check if the stepped thread
has advanced already on hardware step targets.
(currently_stepping): Return true if the thread stepped a
breakpoint.
gdb/testsuite/
2014-11-12 Pedro Alves <palves@redhat.com>
* gdb.arch/i386-bp_permanent.c: New file.
* gdb.arch/i386-bp_permanent.exp: Don't skip on x86_64.
(srcfile): Set to i386-bp_permanent.c.
(top level): Adjust to work in both 32-bit and 64-bit modes. Test
that stepi does not execute the 'leave' instruction, instead of
testing it does execute.
* gdb.base/bp-permanent.c: New file.
* gdb.base/bp-permanent.exp: New file.
"permanent"-ness is currently a property of the breakpoint. But, it
should actually be an implementation detail of a _location_. Consider
this bit in infrun.c:
/* Normally, by the time we reach `resume', the breakpoints are either
removed or inserted, as appropriate. The exception is if we're sitting
at a permanent breakpoint; we need to step over it, but permanent
breakpoints can't be removed. So we have to test for it here. */
if (breakpoint_here_p (aspace, pc) == permanent_breakpoint_here)
{
if (gdbarch_skip_permanent_breakpoint_p (gdbarch))
gdbarch_skip_permanent_breakpoint (gdbarch, regcache);
else
error (_("\
The program is stopped at a permanent breakpoint, but GDB does not know\n\
how to step past a permanent breakpoint on this architecture. Try using\n\
a command like `return' or `jump' to continue execution."));
}
This will wrongly skip a non-breakpoint instruction if we have a
multiple location breakpoint where the whole breakpoint was set to
"permanent" because one of the locations happened to be permanent,
even if the one GDB is resuming from is not.
Related, because the permanent breakpoints are only marked as such in
init_breakpoint_sal, we currently miss marking momentary breakpoints
as permanent. A test added by a following patch trips on that.
Making permanent-ness be per-location, and marking locations as such
in add_location_to_breakpoint, the natural place to do this, fixes
this issue...
... and then exposes a latent issue with mark_breakpoints_out. It's
clearing the inserted flag of permanent breakpoints. This results in
assertions failing like this:
Breakpoint 1, main () at testsuite/gdb.base/callexit.c:32
32 return 0;
(gdb) call callexit()
[Inferior 1 (process 15849) exited normally]
gdb/breakpoint.c:12854: internal-error: allegedly permanent breakpoint is not actually inserted
A problem internal to GDB has been detected,
further debugging may prove unreliable.
The call dummy breakpoint, which is a momentary breakpoint, is set on
top of a manually inserted breakpoint instruction, and so is now
rightfully marked as a permanent breakpoint. See "Write a legitimate
instruction at the point where the infcall breakpoint is going to be
inserted." comment in infcall.c.
Re. make_breakpoint_permanent. That's only called by solib-pa64.c.
Permanent breakpoints were actually originally invented for HP-UX [1].
I believe that that call (the only one in the tree) is unnecessary
nowadays, given that nowadays the core breakpoints code analyzes the
instruction under the breakpoint to automatically detect whether it's
setting a breakpoint on top of a breakpoint instruction in the
program. I know close to nothing about HP-PA/HP-UX, though.
[1] https://sourceware.org/ml/gdb-patches/1999-q3/msg00245.html, and
https://sourceware.org/ml/gdb-patches/1999-q3/msg00242.html
In addition to the per-location issue, "permanent breakpoints" are
currently always displayed as enabled=='n':
(gdb) b main
Breakpoint 3 at 0x40053c: file ../../../src/gdb/testsuite/gdb.arch/i386-permbkpt.S, line 29.
(gdb) info breakpoints
Num Type Disp Enb Address What
3 breakpoint keep n 0x000000000040053c ../../../src/gdb/testsuite/gdb.arch/i386-permbkpt.S:29
But OTOH they're always enabled; there's no way to disable them...
In turn, this means that if one adds commands to such a breakpoint,
they're _always_ run:
(gdb) start
Starting program: /home/pedro/gdb/mygit/build/gdb/testsuite/gdb.arch/i386-permbkpt
...
Temporary breakpoint 1, main () at ../../../src/gdb/testsuite/gdb.arch/i386-permbkpt.S:29
29 int3
(gdb) b main
Breakpoint 2 at 0x40053c: file ../../../src/gdb/testsuite/gdb.arch/i386-permbkpt.S, line 29.
(gdb) info breakpoints
Num Type Disp Enb Address What
2 breakpoint keep n 0x000000000040053c ../../../src/gdb/testsuite/gdb.arch/i386-permbkpt.S:29
(gdb) commands
Type commands for breakpoint(s) 2, one per line.
End with a line saying just "end".
>echo "hello!"
>end
(gdb) disable 2
(gdb) start
The program being debugged has been started already.
Start it from the beginning? (y or n) y
Temporary breakpoint 3 at 0x40053c: file ../../../src/gdb/testsuite/gdb.arch/i386-permbkpt.S, line 29.
Starting program: /home/pedro/gdb/mygit/build/gdb/testsuite/gdb.arch/i386-permbkpt
Breakpoint 2, main () at ../../../src/gdb/testsuite/gdb.arch/i386-permbkpt.S:29
29 int3
"hello!"(gdb)
IMO, one should be able to disable such a breakpoint, and GDB should
then behave just like if the user hadn't created the breakpoint in the
first place (that is, report a SIGTRAP).
By making permanent-ness a property of the location, and eliminating
the bp_permanent enum enable_state state ends up fixing that as well.
No tests are added for these changes yet; they'll be added in a follow
up patch, as skipping permanent breakpoints is currently broken and
trips on an assertion in infrun.
Tested on x86_64 Fedora 20, native and gdbserver.
gdb/ChangeLog:
2014-11-12 Pedro Alves <palves@redhat.com>
Mark locations as permanent, not the whole breakpoint.
* breakpoint.c (remove_breakpoint_1, remove_breakpoint): Adjust.
(mark_breakpoints_out): Don't mark permanent breakpoints as
uninserted.
(breakpoint_init_inferior): Use mark_breakpoints_out.
(breakpoint_here_p): Adjust.
(bpstat_stop_status, describe_other_breakpoints): Remove handling
of permanent breakpoints.
(make_breakpoint_permanent): Mark each location as permanent,
instead of marking the breakpoint.
(add_location_to_breakpoint): If the location is permanent, mark
it as such, and as inserted.
(init_breakpoint_sal): Don't make the breakpoint permanent here.
(bp_location_compare, update_global_location_list): Adjust.
(update_breakpoint_locations): Don't make the breakpoint permanent
here.
(disable_breakpoint, enable_breakpoint_disp): Don't skip permanent
breakpoints.
* breakpoint.h (enum enable_state) <bp_permanent>: Delete field.
(struct bp_location) <permanent>: New field.
* guile/scm-breakpoint.c (bpscm_enable_state_to_string): Remove
reference to bp_permanent.
breakpoint.c uses gdbarch_breakpoint_from_pc to determine whether a
breakpoint location points at a permanent breakpoint:
static int
bp_loc_is_permanent (struct bp_location *loc)
{
...
addr = loc->address;
bpoint = gdbarch_breakpoint_from_pc (loc->gdbarch, &addr, &len);
...
if (target_read_memory (loc->address, target_mem, len) == 0
&& memcmp (target_mem, bpoint, len) == 0)
retval = 1;
...
So I think we should default the gdbarch_skip_permanent_breakpoint
hook to advancing the PC by the length of the breakpoint instruction,
as determined by gdbarch_breakpoint_from_pc. I believe that simple
implementation does the right thing for most architectures. If
there's an oddball architecture where that doesn't work, then it
should override the hook, just like it should be overriding the hook
if there was no default anyway.
The only two implementation of skip_permanent_breakpoint are
i386_skip_permanent_breakpoint, for x86, and
hppa_skip_permanent_breakpoint, for PA-RISC/HP-UX
The x86 implementation is trivial, and can clearly be replaced by the
new default.
I don't know about the HP-UX one though, I know almost nothing about
PA. It may well be advancing the PC ends up being equivalent.
Otherwise, it must be that "jump $pc_after_bp" doesn't work either...
Tested on x86_64 Fedora 20 native and gdbserver.
gdb/
2014-11-12 Pedro Alves <palves@redhat.com>
* arch-utils.c (default_skip_permanent_breakpoint): New function.
* arch-utils.h (default_skip_permanent_breakpoint): New
declaration.
* gdbarch.sh (skip_permanent_breakpoint): Now an 'f' function.
Install default_skip_permanent_breakpoint as default method.
* i386-tdep.c (i386_skip_permanent_breakpoint): Delete function.
(i386_gdbarch_init): Don't install it.
* infrun.c (resume): Assume there's always a
gdbarch_skip_permanent_breakpoint implementation.
* gdbarch.h, gdbarch.c: Regenerate.
