G++ in GCC 4.2 silently ignores
---
__attribute__ ((section(".rodata.v1_a2")))
const short rodata_item1 = 101;
---
which leads to plugin_layout_with_alignment test failure with
Expected 12 sections, found 8 sections
since 4 .rodata sections are missing. As a workaround, this patch
changes plugin_layout_with_alignment test from C++ to C.
* testsuite/plugin_layout_with_alignment.cc: Renamed to ..
* testsuite/plugin_layout_with_alignment.c: This.
* testsuite/Makefile.am (plugin_layout_with_alignment.o): Updated.
(plugin_layout_with_alignment): Likewise.
* testsuite/Makefile.in: Regenerated.
Functions compiled with the gcc option `-mhotpatch' may start with a
branch-never BRCL instruction as a 6-byte NOP. And functions compiled
with `-mstack-size' contain a BRC instruction in their prologue that is
actually a conditional trap. Both of these special jumps cause the
prologue parser to stop and yield bad unwinding results.
This change makes the prologue analyzer recognize such special jumps and
ignore them.
gdb/ChangeLog:
* s390-linux-tdep.c (s390_analyze_prologue): Ignore BRC and BRCL
instructions that do nothing or are conditional traps.
When determining the frame ID of an inline frame, GDB currently asserts
that a valid ID of the underlying real frame is found, and that it does
not match outer_frame_id. From inline_frame_this_id():
/* For now, require we don't match outer_frame_id either (see
comment above). */
gdb_assert (!frame_id_eq (*this_id, outer_frame_id));
However, this assertion may fail when the real frame's unwinder can not
determine the frame ID. This happened on an s390x target with a binary
that lacked call frame information and also confused the prologue
analyzer, because then s390_frame_this_id() left the frame ID at its
default.
To fix this, this change enhances s390_frame_this_id such that an
unavailable-stack frame ID is built if no frame base can be determined
but the function address is available.
gdb/ChangeLog:
* s390-linux-tdep.c (s390_prologue_frame_unwind_cache): Store
frame func's PC in info->func before any other failure can occur.
(s390_frame_this_id): Use frame_id_build_unavailable_stack if
info->func has been filled out.
It's not possible today to select some of the osabis by name.
Specifically, those that have spaces in their names and then the first
word is ambiguous...
For example:
(gdb) set osabi <TAB>
[...]
FreeBSD ELF
FreeBSD a.out
[...]
(gdb) set osabi FreeBSD ELF
Ambiguous item "FreeBSD ELF".
In reality, because "set osabi" is an enum command, that was
equivalent to trying "set osabi FreeBSD", which is then obviously
ambiguous, because of "FreeBSD ELF" and "FreeBSD a.out".
Also, even if the first word is not ambiguous, we actually ignore
whatever comes after the first word:
(gdb) set osabi GNU/Linux
(gdb) show osabi
The current OS ABI is "GNU/Linux".
The default OS ABI is "GNU/Linux".
(gdb) set osabi Windows SomeNonsense
^^^^^^^^^^^^
(gdb) show osabi
The current OS ABI is "Windows CE".
The default OS ABI is "GNU/Linux".
(gdb)
Fix this by avoiding spaces in osabi names.
We could instead make "set osabi" have a custom set hook, or
alternatively make the enum set hook (in cli-setshow.c) handle values
with spaces, but OTOH, I have a feeling that could cause trouble.
E.g., in cases where we might want to write more than one enum value
in the same line. We could support quoting as workaround, but, not
sure we want that. "No spaces" seems like a simpler rule.
gdb/ChangeLog:
2016-03-09 Pedro Alves <palves@redhat.com>
* osabi.c (gdb_osabi_names): Avoid spaces in osabi names.
Even though "set architecture" presents fr300 as option:
(gdb) set architecture fr<TAB>
fr300 fr400 fr450 fr500 fr550 frv
Actually selecting fr300 doesn't work:
(gdb) set architecture fr300
Architecture `fr300' not recognized.
The target architecture is set automatically (currently i386)
(gdb)
This just looks like an obvious oversight. Looking around gcc and
binutils sources, FR300 is basically a FR500 specialized for DSP and
low power.
gdb/ChangeLog:
2016-03-09 Pedro Alves <palves@redhat.com>
* frv-tdep.c (frv_gdbarch_init): Handle bfd_mach_fr300.
Currently, it's not possible to manually set some of the v850 archs in
gdb:
(gdb) set architecture v850<TAB>
v850 (using old gcc ABI)
v850-rh850
v850e
v850e (using old gcc ABI)
v850e1
[...]
(gdb) set architecture v850 (using old gcc ABI)
Ambiguous item "v850 (using old gcc ABI)".
The problem is that "set architecture" is a GDB "enum command", and
GDB only considers an enum value to be the string up until the first
space. So writing "v850 (using old gcc ABI)" is the same as writing
"v850", and then that's not an unambiguous arch printable name prefix.
v850 is actually the only arch that has spaces in its printable name.
One can conveniently see that with e.g.:
(gdb) set max-completions unlimited
(gdb) complete set architecture
...
Rather than hack GDB into accepting this somehow, make v850 arch
printable names more like the printable names of the other archs, and
put the abi variant in the "machine" part, after a ':'.
We now get:
(gdb) set architecture v850<TAB>
v850:old-gcc-abi
v850:rh850
v850e
v850e1
v850e1:old-gcc-abi
v850e2
v850e2:old-gcc-abi
[...]
And now "set architecture v850:old-gcc-abi" works as expected.
I ran the binutils/gas/ld testsuites, and found no regressions. I
don't have a cross compiler handy, but I ran the gdb tests anyway,
which covers at least some snoke testing.
I think that the OUTPUT_ARCH in ld/scripttempl/v850.sc may have got
broken with the previous 2012 change, since I hacked v850_rh850.sc to
output "v850" and ld failed to grok it. I think it only works if the
old GCC ABI is the configured v850 default ABI. That's now fixed by
changing to use explicit v850:old-gcc-abi.
Also, this actually "fixes" an existing GDB test, which isn't likewise
expecting spaces in arch names, when GDB is configured for
--target=v850:
(gdb) FAIL: gdb.xml/tdesc-arch.exp: read valid architectures
bfd/ChangeLog:
2016-03-09 Pedro Alves <palves@redhat.com>
* cpu-v850.c (N): Append ":old-gcc-abi" instead of " (using old
gcc ABI)" to printable name.
* cpu-v850_rh850.c (bfd_v850_rh850_arch): Use "v850:rh850" instead
of "v850-rh850" as printable name.
ld/ChangeLog:
2016-03-09 Pedro Alves <palves@redhat.com>
* scripttempl/v850.sc: Use "v850:old-gcc-abi" as OUTPUT_ARCH.
* scripttempl/v850_rh850.sc: Use "v850:rh850" as OUTPUT_ARCH.
gcc 4.7 complains about variables that shadow function names, which now happens
in tc-arm.c because there is a global function do_align (), and local variables
do_align. The simplest fix for this seems to be to rename those variables to
do_alignment.
gas/ChangeLog:
2016-03-09 Trevor Saunders <tbsaunde+binutils@tbsaunde.org>
* config/tc-arm.c (neon_alignment_bit): Rename do_align to
do_alignment.
(do_neon_ld_st_lane): Likewise.
(do_neon_ld_dup): Likewise.
Since dynamic sections aren't applicable to relocatable file, don't
create dynamic sections for -E/--dynamic-list when relocatable.
bfd/
PR ld/19789
* elflink.c (elf_link_add_object_symbols): Create dynamic sections
for -E/--dynamic-list only when not relocatable.
ld/
PR ld/19789
* testsuite/ld-elf/pr19789.d: New file.
* testsuite/ld-elf/pr19789.s: Likewise.
GCC 6 does not generate constructors for two of gold's test cases.
This patch simply removes the checks for them.
gold/
PR 19751
* testsuite/Makefile.am (retain_symbols_file_test): Remove check
for constructor.
* testsuite/Makefile.in: Regenerate.
* testsuite/dynamic_list.sh: Likewise.
* testsuite/retain_symbols_file_test.sh: Likewise.
This fixes:
$ ./gdb -q -ex "set endian big" -ex "set architecture cris"
The target is assumed to be big endian
.../src/gdb/cris-tdep.c:4051: internal-error: cris_gdbarch_init: big endian byte order in info
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n)
The "set cris-version" command can likewise cause internal errors.
The gdbarch init routine should be returning 0 to reject the
architecture instead of internal erroring on user input.
gdb/ChangeLog:
2016-03-09 Pedro Alves <palves@redhat.com>
* cris-tdep.c (cris_gdbarch_init): Return 0 if the info's byte
order is BFD_ENDIAN_BIG or if the cris version is unsupported.
Running the testsuite with a gdb configured with --enable-libmcheck
reveals a problem:
(gdb) ptype 3 * 2.0
type = <12-byte float>
memory clobbered past end of allocated block
ERROR: Process no longer exists
UNRESOLVED: gdb.ada/ptype_arith_binop.exp: ptype 3 * 2.0
(gdb) PASS: gdb.dlang/expression.exp: ptype 0x1.FFFFFFFFFFFFFp1023
ptype 0x1p-52L
type = real
memory clobbered past end of allocated block
ERROR: Process no longer exists
UNRESOLVED: gdb.dlang/expression.exp: ptype 0x1p-52L
Even though this shows up with Ada and D, it's easy to reproduce in C
too. We just need to print a long double, when the current arch is
32-bit, which is the default when gdb starts up:
$ ./gdb -q -ex "ptype 1.0L"
type = long double
memory clobbered past end of allocated block
Aborted (core dumped)
Valgrind shows:
==22159== Invalid write of size 8
==22159== at 0x8464A9: floatformat_from_doublest (doublest.c:756)
==22159== by 0x846822: store_typed_floating (doublest.c:867)
==22159== by 0x6A7959: value_from_double (value.c:3662)
==22159== by 0x6A9F2D: evaluate_subexp_standard (eval.c:745)
==22159== by 0x7F31AF: evaluate_subexp_c (c-lang.c:716)
==22159== by 0x6A8986: evaluate_subexp (eval.c:79)
==22159== by 0x6A8BA3: evaluate_type (eval.c:174)
==22159== by 0x817CCF: whatis_exp (typeprint.c:456)
==22159== by 0x817EAA: ptype_command (typeprint.c:508)
==22159== by 0x5F267B: do_cfunc (cli-decode.c:105)
==22159== by 0x5F5618: cmd_func (cli-decode.c:1885)
==22159== by 0x83622A: execute_command (top.c:475)
==22159== Address 0x8c6cb28 is 8 bytes inside a block of size 12 alloc'd
==22159== at 0x4C2AA98: calloc (vg_replace_malloc.c:711)
==22159== by 0x87384A: xcalloc (common-utils.c:83)
==22159== by 0x873889: xzalloc (common-utils.c:93)
==22159== by 0x6A34CB: allocate_value_contents (value.c:1036)
==22159== by 0x6A3501: allocate_value (value.c:1047)
==22159== by 0x6A790A: value_from_double (value.c:3656)
==22159== by 0x6A9F2D: evaluate_subexp_standard (eval.c:745)
==22159== by 0x7F31AF: evaluate_subexp_c (c-lang.c:716)
==22159== by 0x6A8986: evaluate_subexp (eval.c:79)
==22159== by 0x6A8BA3: evaluate_type (eval.c:174)
==22159== by 0x817CCF: whatis_exp (typeprint.c:456)
==22159== by 0x817EAA: ptype_command (typeprint.c:508)
==22159==
type = long double
(gdb)
Even if the target and host floating-point formats match, the length
of the types might still be different. On x86, long double is the
80-bit extended precision type on both 32-bit and 64-bit ABIs, but by
default it is stored as 12 bytes on 32-bit, and 16 bytes on 64-bit,
for alignment reasons. Several places in doublest.c already consider
this, but floatformat_to_doublest and floatformat_from_doublest miss
it. E.g., convert_typed_floating and store_typed_floating,
Tested on x86-64 Fedora 23 with --enable-libmcheck, where it fixes the
crashed above.
gdb/ChangeLog:
2016-03-09 Pedro Alves <palves@redhat.com>
* doublest.c: Extend comments.
(floatformat_to_doublest, floatformat_from_doublest): Copy the
floatformat's total size, not the host type's size.
This would have caught the HP/PA bug fixed in the previous patch:
.../src/gdb/gdbtypes.c:4690: internal-error: arch_float_type: Assertion `len >= floatformat_totalsize_bytes (floatformats[0])' failed.
A problem internal to GDB has been detected,
further debugging may prove unreliable.
Quit this debugging session? (y or n)
Tested on x86-64 Fedora 23, --enable-targets=all.
gdb/ChangeLog:
2016-03-09 Pedro Alves <palves@redhat.com>
* doublest.c (floatformat_totalsize_bytes): New function.
(floatformat_from_type): Assert that the type's length is at least
as long as the floatformat's totalsize.
* doublest.h (floatformat_totalsize_bytes): New declaration.
* gdbtypes.c (arch_float_type): Assert that the type's length is
at least as long as the floatformat's totalsize.
This:
$ ./gdb -ex "set architecture hppa1.0" -ex "set osabi GNU/Linux" -ex "ptype 1.0L"
Shows that HPPA/Linux support for long doubles is broken. It causes
GDB to access memory out of bounds. With Valgrind, we see:
The target architecture is assumed to be hppa1.0
==4371== Invalid write of size 8
==4371== at 0x4C2F21F: memset (vg_replace_strmem.c:1224)
==4371== by 0x8451C4: convert_doublest_to_floatformat (doublest.c:362)
==4371== by 0x845F86: floatformat_from_doublest (doublest.c:769)
==4371== by 0x84628E: store_typed_floating (doublest.c:873)
==4371== by 0x6A7C3D: value_from_double (value.c:3662)
==4371== by 0x6AA211: evaluate_subexp_standard (eval.c:745)
==4371== by 0x7F306D: evaluate_subexp_c (c-lang.c:716)
==4371== by 0x6A8C6A: evaluate_subexp (eval.c:79)
==4371== by 0x6A8E87: evaluate_type (eval.c:174)
==4371== by 0x817B8D: whatis_exp (typeprint.c:456)
==4371== by 0x817D68: ptype_command (typeprint.c:508)
==4371== by 0x5F2977: do_cfunc (cli-decode.c:105)
==4371== Address 0x8998d18 is 0 bytes after a block of size 8 alloc'd
==4371== at 0x4C2AA98: calloc (vg_replace_malloc.c:711)
==4371== by 0x8732B6: xcalloc (common-utils.c:83)
==4371== by 0x8732F5: xzalloc (common-utils.c:93)
==4371== by 0x6A37AF: allocate_value_contents (value.c:1036)
==4371== by 0x6A37E5: allocate_value (value.c:1047)
==4371== by 0x6A7BEE: value_from_double (value.c:3656)
==4371== by 0x6AA211: evaluate_subexp_standard (eval.c:745)
==4371== by 0x7F306D: evaluate_subexp_c (c-lang.c:716)
==4371== by 0x6A8C6A: evaluate_subexp (eval.c:79)
==4371== by 0x6A8E87: evaluate_type (eval.c:174)
==4371== by 0x817B8D: whatis_exp (typeprint.c:456)
==4371== by 0x817D68: ptype_command (typeprint.c:508)
The trouble is that hppa_linux_init_abi overrides the default
long_double_bit set by the generic hppa-tdep.c:
set_gdbarch_long_double_bit (gdbarch, 128);
set_gdbarch_long_double_format (gdbarch, floatformats_ia64_quad);
with:
/* On hppa-linux, currently, sizeof(long double) == 8. There has been
some discussions to support 128-bit long double, but it requires some
more work in gcc and glibc first. */
set_gdbarch_long_double_bit (gdbarch, 64);
which misses overriding the long_double_format, so we end with a weird
combination of:
set_gdbarch_long_double_bit (gdbarch, 64);
set_gdbarch_long_double_format (gdbarch, floatformats_ia64_quad);
Weird because floatformats_ia64_quad's totalsize is longer than 64-bits.
The floatformat conversion routines use the struct floatformat's
totalsize (in bits) to know how much to copy/convert, thus the buffer
overruns.
gdb/ChangeLog:
2016-03-09 Pedro Alves <palves@redhat.com>
* hppa-linux-tdep.c (hppa_linux_init_abi): Set the long double
format to floatformats_ieee_double.
This patch is a simple refactoring that will allow the MIPS backend to
replace the Output_data_reloc_base::do_write() method without copying
its entire implementation. I've moved the implementation of do_write()
into a function template, which can be instantiated with a custom
class to write the MIPS-specific relocation format. The custom class
for MIPS needs access to the symbol index and address from
Output_reloc, so I've included the part of Vlad's MIPS-64 patch that
makes those accessor methods public.
2016-03-08 Cary Coutant <ccoutant@gmail.com>
Vladimir Radosavljevic <vladimir.radosavljevic@imgtec.com>
gold/
* output.cc (Output_reloc_writer): New type.
(Output_data_reloc_base::do_write): Move implementation to template
in output.h and replace with invocation of template.
* output.h (Output_file): Move to top of file.
(Output_reloc::get_symbol_index): Move to public interface.
(Output_reloc::get_address): Likewise.
(Output_data_reloc_base::do_write_generic): New function template.
Increment PLT reference count for locally defined local IFUNC symbols
in shared object since STT_GNU_IFUNC symbol must go through PLT even
if it is locally defined and undefined symbol may turn out to be a
STT_GNU_IFUNC symbol later.
bfd/
PR ld/19784
* elf32-i386.c (elf_i386_check_relocs): Increment PLT reference
count for locally defined local IFUNC symbols in shared object.
* elf64-x86-64.c (elf_x86_64_check_relocs): Likewise.
ld/
PR ld/19784
* testsuite/ld-i386/i386.exp: Remove pr19636-2e-nacl test.
* testsuite/ld-i386/pr19636-2e-nacl.d: Moved to ...
* testsuite/ld-i386/pr19636-2e.d: Here. Remove notarget.
* testsuite/ld-ifunc/ifunc.exp: Run PR ld/19784 tests.
* testsuite/ld-ifunc/pass.out: New file.
* testsuite/ld-ifunc/pr19784a.c: Likewise.
* testsuite/ld-ifunc/pr19784b.c: Likewise.
* testsuite/ld-ifunc/pr19784c.c: Likewise.
Since compiler may pass --as-needed to ld by default, link .o file
before .so file in ifunc tests.
PR ld/19774
* testsuite/ld-ifunc/ifunc.exp: Link tmpdir/pr18808a.o before
tmpdir/libpr18808.so. Link tmpdir/pr18841a.o before
tmpdir/libpr18841b.so and tmpdir/libpr18841c.so. Test
--as-needed for pr18841c.
bfd/
2016-03-08 Cupertino Miranda <Cupertino.Miranda@synopsys.com>
Andrew Burgess <andrew.burgess@embecosm.com>
* elf32-arc.c (arc_bfd_get_32): Becomes an alias for bfd_get_32.
(arc_bfd_put_32): Becomes an alias for bfd_put_32.
(arc_elf_howto_init): Added assert to validate relocations.
(get_middle_endian_relocation): Delete.
(middle_endian_convert): New function.
(ME): Redefine, now does nothing.
(IS_ME): New define.
(arc_do_relocation): Extend the attached 'ARC_RELOC_HOWTO'
definition to call middle_endian_convert. Add a new local
variable and make use of this throughout. Added call to
arc_bfd_get_8 and arc_bfd_put_8 for 8 bit relocations.
gas/
2016-03-08 Andrew Burgess <andrew.burgess@embecosm.com>
* testsuite/gas/arc/inline-data-1.d: New file.
* testsuite/gas/arc/inline-data-1.s: New file.
include/
2016-03-08 Cupertino Miranda <Cupertino.Miranda@synopsys.com>
Andrew Burgess <andrew.burgess@embecosm.com>
* elf/arc-reloc.def: Add a call to ME within the formula for each
relocation that requires middle-endian correction.
bfd PR binutils/19775
* archive.c (bfd_generic_openr_next_archived_file): Allow zero
length elements in the archive.
binutils PR binutils/19775
* testsuite/binutils-all/ar.exp (proc empty_archive): New proc.
Run the new proc.
* testsuite/binutils-all/empty: New, empty, file.
Since compiler may pass --as-needed to ld by default, link .o file
before .so file in i386/x86-64 tests.
PR ld/19774
* testsuite/ld-i386/i386.exp: Link tmpdir/pr18900.o before
tmpdir/pr18900.so and test --as-needed. Link tmpdir/gotpc1.o
before tmpdir/got1d.so and test --as-needed.
* testsuite/ld-x86-64/x86-64.exp: Link tmpdir/pr18900.o before
tmpdir/pr18900.so and test --as-needed.
Fix this GDB crash:
$ gdb -ex "set architecture mips:10000"
Segmentation fault (core dumped)
Backtrace:
Program received signal SIGSEGV, Segmentation fault.
0x0000000000495b1b in mips_gdbarch_init (info=..., arches=0x0) at /home/pedro/gdb/mygit/cxx-convertion/src/gdb/mips-tdep.c:8436
8436 if (bfd_get_flavour (info.abfd) == bfd_target_elf_flavour
(top-gdb) bt
#0 0x0000000000495b1b in mips_gdbarch_init (info=..., arches=0x0) at .../src/gdb/mips-tdep.c:8436
#1 0x00000000007348a6 in gdbarch_find_by_info (info=...) at .../src/gdb/gdbarch.c:5155
#2 0x000000000073563c in gdbarch_update_p (info=...) at .../src/gdb/arch-utils.c:522
#3 0x0000000000735585 in set_architecture (ignore_args=0x0, from_tty=1, c=0x26bc870) at .../src/gdb/arch-utils.c:496
#4 0x00000000005f29fd in do_sfunc (c=0x26bc870, args=0x0, from_tty=1) at .../src/gdb/cli/cli-decode.c:121
#5 0x00000000005fd3f3 in do_set_command (arg=0x7fffffffdcdd "mips:10000", from_tty=1, c=0x26bc870) at .../src/gdb/cli/cli-setshow.c:455
#6 0x0000000000836157 in execute_command (p=0x7fffffffdcdd "mips:10000", from_tty=1) at .../src/gdb/top.c:460
#7 0x000000000071abfb in catch_command_errors (command=0x835f6b <execute_command>, arg=0x7fffffffdccc "set architecture mips:10000", from_tty=1)
at .../src/gdb/main.c:368
#8 0x000000000071bf4f in captured_main (data=0x7fffffffd750) at .../src/gdb/main.c:1132
#9 0x0000000000716737 in catch_errors (func=0x71af44 <captured_main>, func_args=0x7fffffffd750, errstring=0x106b9a1 "", mask=RETURN_MASK_ALL)
at .../src/gdb/exceptions.c:240
#10 0x000000000071bfe6 in gdb_main (args=0x7fffffffd750) at .../src/gdb/main.c:1164
#11 0x000000000040a6ad in main (argc=4, argv=0x7fffffffd858) at .../src/gdb/gdb.c:32
(top-gdb)
We already check whether info.abfd is NULL before all other
bfd_get_flavour calls in the same function. Just this one case was
missing.
(This was exposed by a WIP test that tries all "set architecture ARCH"
values.)
gdb/ChangeLog:
2016-03-07 Pedro Alves <palves@redhat.com>
* mips-tdep.c (mips_gdbarch_init): Check whether info.abfd is NULL
before calling bfd_get_flavour.
2016-03-07 Jiong Wang <jiong.wang@arm.com>
bfd/
* elfnn-aarch64.c (elfNN_aarch64_check_relocs): Always create .got section
if the symbol "_GLOBAL_OFFSET_TABLE_" referenced.
ld/
* testsuite/ld-aarch64/implicit_got_section_1.s: New test source file.
* testsuite/ld-aarch64/implicit_got_section_1.d: New test expected result.
* testsuite/ld-aarch64/aarch64-elf.exp: Run new test.
I forgot to do it in my previous commit. This is necessary because we
execute the script directly on gdb/testsuite/Makefile.in.
gdb/testsuite/ChangeLog:
2016-03-06 Sergio Durigan Junior <sergiodj@redhat.com>
* analyze-racy-logs.py: Set executable bit.
Since compiler may pass --as-needed to ld by default, link
tmpdir/copyreloc-main.o before tmpdir/copyreloc-lib.so.
* testsuite/ld-i386/i386.exp: Link tmpdir/copyreloc-main.o
before tmpdir/copyreloc-lib.so and test --as-needed.
* testsuite/ld-x86-64/x86-64.exp: Likewise.
This is an initial attempt to introduce some mechanisms to identify
racy testcases present in our testsuite. As can be seen in previous
discussions, racy tests are really bothersome and cause our BuildBot
to pollute the gdb-testers mailing list with hundreds of
false-positives messages every month. Hopefully, identifying these
racy tests in advance (and automatically) will contribute to the
reduction of noise traffic to gdb-testers, maybe to the point where we
will be able to send the failure messages directly to the authors of
the commits.
I spent some time trying to decide the best way to tackle this
problem, and decided that there is no silver bullet. Racy tests are
tricky and it is difficult to catch them, so the best solution I could
find (for now?) is to run our testsuite a number of times in a row,
and then compare the results (i.e., the gdb.sum files generated during
each run). The more times you run the tests, the more racy tests you
are likely to detect (at the expense of waiting longer and longer).
You can also run the tests in parallel, which makes things faster (and
contribute to catching more racy tests, because your machine will have
less resources for each test and some of them are likely to fail when
this happens). I did some tests in my machine (8-core i7, 16GB RAM),
and running the whole GDB testsuite 5 times using -j6 took 23 minutes.
Not bad.
In order to run the racy test machinery, you need to specify the
RACY_ITER environment variable. You will assign a number to this
variable, which represents the number of times you want to run the
tests. So, for example, if you want to run the whole testsuite 3
times in parallel (using 2 cores), you will do:
make check RACY_ITER=3 -j2
It is also possible to use the TESTS variable and specify which tests
you want to run:
make check TEST='gdb.base/default.exp' RACY_ITER=3 -j2
And so on. The output files will be put at the directory
gdb/testsuite/racy_outputs/.
After make invokes the necessary rules to run the tests, it finally
runs a Python script that will analyze the resulting gdb.sum files.
This Python script will read each file, and construct a series of sets
based on the results of the tests (one set for FAIL's, one for
PASS'es, one for KFAIL's, etc.). It will then do some set operations
and come up with a list of unique, sorted testcases that are racy.
The algorithm behind this is:
for state in PASS, FAIL, XFAIL, XPASS...; do
if a test's state in every sumfile is $state; then
it is not racy
else
it is racy
(The algorithm is actually a bit more complex than that, because it
takes into account other things in order to decide whether the test
should be ignored or not).
IOW, a test must have the same state in every sumfile.
After processing everything, the script prints the racy tests it could
identify on stdout. I am redirecting this to a file named racy.sum.
Something else that I wasn't sure how to deal with was non-unique
messages in our testsuite. I decided to do the same thing I do in our
BuildBot: include a unique identifier in the end of message, like:
gdb.base/xyz.exp: non-unique message
gdb.base/xyz.exp: non-unique message <<2>>
This means that you will have to be careful about them when you use
the racy.sum file.
I ran the script several times here, and it did a good job catching
some well-known racy tests. Overall, I am satisfied with this
approach and I think it will be helpful to have it upstream'ed. I
also intend to extend our BuildBot and create new, specialized
builders that will be responsible for detecting the racy tests every X
number of days.
2016-03-05 Sergio Durigan Junior <sergiodj@redhat.com>
* Makefile.in (DEFAULT_RACY_ITER): New variable.
(CHECK_TARGET_TMP): Likewise.
(check-single-racy): New rule.
(check-parallel-racy): Likewise.
(TEST_TARGETS): Adjust rule to account for RACY_ITER.
(do-check-parallel-racy): New rule.
(check-racy/%.exp): Likewise.
* README (Racy testcases): New section.
* analyze-racy-logs.py: New file.
When calling function with argument of size more than 8 bytes fails with
an error "That operation is not available on integers of more than 8 bytes.".
avr-gdb considers only 8 bytes (sizeof(long long)) in case of passing the
argument in registers. When the argument is of size more than 8 byte
then the utility function to extract bytes failed with the above error.
gdb/
* avr-tdep.c (AVR_LAST_ARG_REGNUM): Define.
(avr_push_dummy_call): Correct last needed argument register.
Write MSB of argument into register and subsequent bytes into
other registers in decreasing order.
ARM process record gets the wrong register number for VMOV (from core
register to single-precision register). That is, we should record
the D register rather than the S pseudo register. The patch also
removes the condition "bit (arm_insn_r->arm_insn, 20)" check, which
has been checked above.
It fixes the following internal error,
(gdb) PASS: gdb.reverse/finish-precsave.exp: BP at end of main
continue^M
Continuing.^M
../../binutils-gdb/gdb/regcache.c:649: internal-error: regcache_raw_read: Assertion `regnum >= 0 && regnum < regcache->descr->nr_raw_registers' failed.^M
A problem internal to GDB has been detected,FAIL: gdb.reverse/finish-precsave.exp: run to end of main (GDB internal error)
gdb:
2016-03-04 Yao Qi <yao.qi@linaro.org>
* arm-tdep.c (arm_record_vdata_transfer_insn): Simplify the
condition check. Record the right D register number.
This patch removes the printing "Process record does not support",
and do the print by calling arm_record_unsupported_insn in the
caller. Also, call arm_record_extension_space only when condition
is 0xf.
gdb:
2016-03-04 Yao Qi <yao.qi@linaro.org>
* arm-tdep.c (arm_record_extension_space): Remove code
printing "Process record does not support".
(arm_record_data_proc_misc_ld_str): Likewise.
(decode_insn): Call arm_record_extension_space if condition
is 0xf. Call arm_record_unsupported_insn if ret isn't
ARM_RECORD_SUCCESS. Use 'ret' instead of 'insn_id' to hold
the value of thumb2_record_decode_insn_handler.
I found that odd that passing no arguments to feature_to_c.sh produces
this:
$ ./feature_to_c.sh
./feature_to_c.sh: 23: shift: can't shift that many
but passing one argument shows the help:
$ ./feature_to_c.sh hello
Usage: ./feature_to_c.sh OUTPUTFILE INPUTFILE...
This patch changes the script to show the help in both cases.
gdb/ChangeLog:
* features/feature_to_c.sh: Print the help when passing no
argument.
Common symbol in executable is a definition, which overrides definition
from shared objects. When linker sees a new definition from a shared
object, the new dynamic definition should be overridden by the previous
common symbol in executable.
bfd/
PR ld/19579
* elflink.c (_bfd_elf_merge_symbol): Treat common symbol in
executable as definition if the new definition comes from a
shared library.
ld/
PR ld/19579
* testsuite/ld-elf/pr19579a.c: New file.
* testsuite/ld-elf/pr19579b.c: Likewise.
* testsuite/ld-elf/shared.exp: Run PR ld/19579 test.
gold/
PR gold/19019
* layout.h (Layout::add_target_specific_dynamic_tag): New function.
* layout.cc (Layout::add_target_specific_dynamic_tag): New function.
* mips.cc (Target_mips::make_symbol): Adjust function signature.
* sparc.cc (Target_sparc::Target_sparc): Initialize register_syms_.
(Target_sparc::do_is_defined_by_abi): Remove test for
STT_SPARC_REGISTER.
(Target_sparc::Register_symbol): New struct type.
(Target_sparc::register_syms_): New data member.
(Target_sparc<64, true>::sparc_info): Set has_make_symbol to true.
(Target_sparc::make_symbol): New function.
(Target_sparc::do_finalize_sections): Add register symbols and new
dynamic table entries.
* symtab.h (Sized_symbol::init_undefined): Add value parameter.
(Symbol_table::add_target_global_symbol): New function.
(Symbol_table::target_symbols_): New data member.
* symtab.cc (Sized_symbol::init_undefined): Add value parameter.
(Symbol_table::Symbol_table): Initialize target_symbols_.
(Symbol_table::add_from_object): Pass additional parameters to
Target::make_symbol.
(Symbol_table::define_special_symbol): Likewise.
(Symbol_table::add_undefined_symbol_from_command_line): Pass 0 for
undefined symbol value.
(Symbol_table::set_dynsym_indexes): Process target-specific symbols.
(Symbol_table::sized_finalize): Likewise.
(Symbol_table::sized_write_globals): Likewise.
* target.h (Sized_target::make_symbol): Add name, st_type, object,
st_shndx, and value parameters.
