This commit is the result of running the gdb/copyright.py script,
which automated the update of the copyright year range for all
source files managed by the GDB project to be updated to include
year 2023.
PR symtab/29694 points out a regression caused by the new DWARF
scanner when the cc-with-gdb-index target board is used.
What happens here is that an older version of gdb will make an index
describing the "A" type as:
[737] A: 1 [global, type]
whereas the new gdb says:
[1008] A: 0 [global, type]
Here the old one is correct because the A in CU 0 is just a
declaration without a size:
<1><45>: Abbrev Number: 10 (DW_TAG_structure_type)
<46> DW_AT_name : A
<48> DW_AT_declaration : 1
<48> DW_AT_sibling : <0x6d>
This patch fixes the problem by introducing the idea of a "type
declaration". I think gdb still needs to recurse into these types,
searching for methods, but by marking the type itself as a
declaration, gdb can skip this type during lookups and when writing
the index.
Regression tested on x86-64 using the cc-with-gdb-index board.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29694
When building gdb with -fsanitize-threads and running test-case
gdb.ada/char_enum_unicode.exp, I run into:
...
WARNING: ThreadSanitizer: data race (pid=21301)^M
Write of size 8 at 0x7b2000008080 by main thread:^M
#0 free <null> (libtsan.so.2+0x4c5e2)^M
#1 _dl_close_worker <null> (ld-linux-x86-64.so.2+0x4b7b)^M
#2 convert_between_encodings() charset.c:584^M
...
#21 cooked_index_functions::expand_symtabs_matching() read.c:18606
...
This is fixed by making cooked_index_functions::expand_symtabs_matching wait
for the cooked index finalization to be done.
Tested on x86_64-linux.
https://sourceware.org/bugzilla/show_bug.cgi?id=29311https://sourceware.org/bugzilla/show_bug.cgi?id=29286
After DWARF has been scanned, the cooked index code does a
"finalization" step in a worker thread. This step combines all the
index entries into a single master list, canonicalizes C++ names, and
splits Ada names to synthesize package names.
While this step is run in the background, gdb will wait for the
results in some situations, and it turns out that this step can be
slow. This is PR symtab/29105.
This can be sped up by parallelizing, at a small memory cost. Now
each index is finalized on its own, in a worker thread. The cost
comes from name canonicalization: if a given non-canonical name is
referred to by multiple indices, there will be N canonical copies (one
per index) rather than just one.
This requires changing the users of the index to iterate over multiple
results. However, this is easily done by introducing a new "chained
range" class.
When run on gdb itself, the memory cost seems rather low -- on my
current machine, "maint space 1" reports no change due to the patch.
For performance testing, using "maint time 1" and "file" will not show
correct results. That approach measures "time to next prompt", but
because the patch only affects background work, this shouldn't (and
doesn't) change. Instead, a simple way to make gdb wait for the
results is to set a breakpoint.
Before:
$ /bin/time -f%e ~/gdb/install/bin/gdb -nx -q -batch \
-ex 'break main' /tmp/gdb
Breakpoint 1 at 0x43ec30: file ../../binutils-gdb/gdb/gdb.c, line 28.
2.00
After:
$ /bin/time -f%e ./gdb/gdb -nx -q -batch \
-ex 'break main' /tmp/gdb
Breakpoint 1 at 0x43ec30: file ../../binutils-gdb/gdb/gdb.c, line 28.
0.65
Regression tested on x86-64 Fedora 34.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29105
cooked_index::m_start is unused and can be removed. I think this was
a leftover from a previous approach in the index finalization code,
and then when rewriting it I forgot to remove it.
Tested by rebuilding.
PR build/29110 points out that GDB fails to build on mingw when the
"win32" thread model is in use. It turns out that the Fedora cross
tools using the "posix" thread model, which somehow manages to support
std::future, whereas the win32 model does not.
While looking into this, I found that the configuring with
--disable-threading will also cause a build failure.
This patch fixes this build by introducing a compatibility wrapper for
std::future.
I am not able to test the win32 thread model build, but I'm going to
ask the reporter to try this patch.
Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=29110
My patches yesterday to unify the DWARF index base classes had a bug
-- namely, I did the wholesale dynamic_cast-to-static_cast too hastily
and introduced a crash. This can be seen by trying to add an index to
a file that has an index, or by running a test like gdb-index-cxx.exp
using the cc-with-debug-names.exp target board.
This patch fixes the crash by introducing a new virtual method and
removing some of the static casts.
At AdaCore, we run the internal gdb test suite in several modes,
including one using the .debug_names index. This caught a regression
caused by the new DWARF indexer.
First, the psymtabs-based .debug_names generator was completely wrong.
However, to avoid making the rewrite series even bigger (fixing the
writer will also require rewriting the .debug_names reader), it
attempted to preserve the weirdness.
However, this was not done properly. For example the old writer did
this:
- case STRUCT_DOMAIN:
- return DW_TAG_structure_type;
The new code, instead, simply preserves the actual DWARF tag -- but
this makes future lookups fail, because the .debug_names reader only
looks for DW_TAG_structure_type.
This patch attempts to revert to the old behavior in the writer.
This introduces dwarf_scanner_base, a base class for all the index
readers in the DWARF code. Then, it changes both mapped_index_base
and cooked_index_vector to derive from this new base class.
Building with clang++-14, I see:
CXX dwarf2/cooked-index.o
In file included from /home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.c:21:
/home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.h:172:12: error: explicitly defaulted move constructor is implicitly deleted [-Werror,-Wdefaulted-function-deleted]
explicit cooked_index (cooked_index &&other) = default;
^
/home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.h:225:16: note: move constructor of 'cooked_index' is implicitly deleted because field 'm_storage' has a deleted move constructor
auto_obstack m_storage;
^
/home/smarchi/src/binutils-gdb/gdb/../gdbsupport/gdb_obstack.h:128:28: note: 'auto_obstack' has been explicitly marked deleted here
DISABLE_COPY_AND_ASSIGN (auto_obstack);
^
In file included from /home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.c:21:
/home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.h:174:17: error: explicitly defaulted move assignment operator is implicitly deleted [-Werror,-Wdefaulted-function-deleted]
cooked_index &operator= (cooked_index &&other) = default;
^
/home/smarchi/src/binutils-gdb/gdb/dwarf2/cooked-index.h:225:16: note: move assignment operator of 'cooked_index' is implicitly deleted because field 'm_storage' has a deleted move assignment operator
auto_obstack m_storage;
^
/home/smarchi/src/binutils-gdb/gdb/../gdbsupport/gdb_obstack.h:128:3: note: 'operator=' has been explicitly marked deleted here
DISABLE_COPY_AND_ASSIGN (auto_obstack);
^
/home/smarchi/src/binutils-gdb/gdb/../include/ansidecl.h:425:8: note: expanded from macro 'DISABLE_COPY_AND_ASSIGN'
void operator= (const TYPE &) = delete
^
We explicitly make cooked_index have a default move constructor and
move assignment operator. But it doesn't actually happen because
cooked_index has a field of type auto_obstack, which isn't movable.
We don't actually need cooked_index to be movable at the moment, so
remove those lines.
Change-Id: Ifc1fe3d7d67e3ae1a14363d6c1869936fe80b0a2
After scanning the CUs, the DWARF indexer merges all the data into a
single vector, canonicalizing C++ names as it proceeds. While not
necessarily single-threaded, this process is currently done in just
one thread, to keep memory costs lower.
However, this work is all done without reference to any data outside
of the indexes. This patch improves the apparent performance of GDB
by moving it to the background. All uses of the index are then made
to wait for this process to complete.
In our ongoing example, this reduces the scanning time on gdb itself
to 0.173937 (wall). Recall that before this patch, the time was
0.668923; and psymbol reader does this in 1.598869. That is, at the
end of this series, we see about a 10x speedup.
This parallelizes the new DWARF indexer. The indexer's storage was
designed so that each storage object and each indexer is fully
independent. This setup makes it simple to scan different CUs
independently.
This patch creates a new cooked index storage object per thread, and
then scans a subset of all the CUs in each such thread, using gdb's
existing thread pool.
In the ongoing "gdb gdb" example, this patch reduces the wall time
down to 0.668923, from 0.903534. (Note that the 0.903534 is the time
for the new index -- that is, when the "enable the new index" patch is
rebased to before this one. However, in the final series, that patch
appears toward the end. Hopefully this isn't too confusing.)
This patch introduces the new DWARF index class. It is called
"cooked" to contrast against a "raw" index, which is mapped from disk
without extra effort.
Nothing constructs a cooked index yet. The essential idea here is
that index entries are created via the "add" method; then when all the
entries have been read, they are "finalize"d -- name canonicalization
is performed and the entries are added to a sorted vector.
Entries use the DWARF name (DW_AT_name) or linkage name, not the full
name as is done for partial symbols.
These two facets -- the short name and the deferred canonicalization
-- help improve the performance of this approach. This will become
clear in later patches, when parallelization is added.
Some special code is needed for Ada, because GNAT only emits mangled
("encoded", in the Ada lingo) names, and so we reconstruct the
hierarchical structure after the fact. This is also done in the
finalization phase.
One other aspect worth noting is that the way the "main" function is
found is different in the new code. Currently gdb will notice
DW_AT_main_subprogram, but won't recognize "main" during reading --
this is done later, via explicit symbol lookup. This is done
differently in the new code so that finalization can be done in the
background without then requiring a synchronization to look up the
symbol.
