* [PATCH 1/2] [gdb/macro] Ignore in-file macro definition with 0 line complaint for clang
@ 2024-05-17 10:47 Tom de Vries
2024-05-17 10:47 ` [PATCH 2/2] [gdb/macro] Ignore malformed macro definition " Tom de Vries
0 siblings, 1 reply; 2+ messages in thread
From: Tom de Vries @ 2024-05-17 10:47 UTC (permalink / raw)
To: gdb-patches
When showing complaints for the exec of test-case
gdb.dwarf2/clang-cli-macro.exp, we get:
...
$ gdb -q -batch -iex "set complaints 5" clang-cli-macro -ex "p main"
During symbol reading: \
debug info gives in-file macro definition with zero line 0: ONE 1
$1 = {int ()} 0x4004b7 <main>
...
The readelf output for the .debug_macro section looks like:
...
Contents of the .debug_macro section:
Offset: 0
Version: 5
Offset size: 4
Offset into .debug_line: 0xe3
DW_MACRO_start_file - lineno: 0 filenum: 1
DW_MACRO_define - lineno : 1 macro : TWO 2
DW_MACRO_end_file
DW_MACRO_define - lineno : 0 macro : ONE 1
...
The complaint is that the DW_MACRO_define for ONE both:
- has lineno 0, so it's predefined or specified on the command line, and
- occurs after the first DW_MACRO_start_file.
In commit e7e7469e7a3 ("gdb: Fix issue with Clang CLI macros") we've added a
workaround to accept this style of .debug_macro section, and
gdb.dwarf2/clang-cli-macro.exp is the test-case for the workaround.
Given that we've added the workaround, it doesn't make sense to complain.
The warning is produced using the following condition in
dwarf_decode_macro_bytes:
...
if ((line == 0 && !at_commandline)
|| (line != 0 && at_commandline))
...
using the variable at_commandline which is initially 1, and set to 0 when
encountering the first DW_MACRO_start_file.
The value of this variable doesn't make sense in the context of a clang-style
.debug_macro section.
We could try to fix this by changing the value of the variable to fit with the
clang-style .debug_macro section, but I don't think it's worth the effort.
Simply fix this by for a clang producer opting out of all complaints in the
function that use the at_commandline variable in the condition.
Tested on x86_64-linux.
---
gdb/dwarf2/macro.c | 10 ++++++----
gdb/testsuite/gdb.dwarf2/clang-cli-macro.exp | 6 ++++++
2 files changed, 12 insertions(+), 4 deletions(-)
diff --git a/gdb/dwarf2/macro.c b/gdb/dwarf2/macro.c
index a511d0a3b44..66a40e88d3a 100644
--- a/gdb/dwarf2/macro.c
+++ b/gdb/dwarf2/macro.c
@@ -546,8 +546,9 @@ dwarf_decode_macro_bytes (dwarf2_per_objfile *per_objfile,
line, body);
break;
}
- if ((line == 0 && !at_commandline)
- || (line != 0 && at_commandline))
+ if (((line == 0 && !at_commandline)
+ || (line != 0 && at_commandline))
+ && !producer_is_clang (cu))
complaint (_("debug info gives %s macro %s with %s line %d: %s"),
at_commandline ? _("command-line") : _("in-file"),
is_define ? _("definition") : _("undefinition"),
@@ -648,8 +649,9 @@ dwarf_decode_macro_bytes (dwarf2_per_objfile *per_objfile,
file = read_unsigned_leb128 (abfd, mac_ptr, &bytes_read);
mac_ptr += bytes_read;
- if ((line == 0 && !at_commandline)
- || (line != 0 && at_commandline))
+ if (((line == 0 && !at_commandline)
+ || (line != 0 && at_commandline))
+ && !producer_is_clang (cu))
complaint (_("debug info gives source %d included "
"from %s at %s line %d"),
file, at_commandline ? _("command-line") : _("file"),
diff --git a/gdb/testsuite/gdb.dwarf2/clang-cli-macro.exp b/gdb/testsuite/gdb.dwarf2/clang-cli-macro.exp
index eafb75ad8d9..fdeaaff5afd 100644
--- a/gdb/testsuite/gdb.dwarf2/clang-cli-macro.exp
+++ b/gdb/testsuite/gdb.dwarf2/clang-cli-macro.exp
@@ -85,6 +85,12 @@ if {[prepare_for_testing "failed to prepare" $testfile [list $srcfile $asm_file]
return
}
+set re_result "[string_to_regexp $]$decimal = \[^\r\n\]+"
+
+with_complaints 5 {
+ gdb_test "print main" ^$re_result "no complaints"
+}
+
if {![runto_main]} {
return
}
base-commit: 44fc9616c2e74396395f60c9a601317e4c4c4733
--
2.35.3
^ permalink raw reply [flat|nested] 2+ messages in thread
* [PATCH 2/2] [gdb/macro] Ignore malformed macro definition complaint for clang
2024-05-17 10:47 [PATCH 1/2] [gdb/macro] Ignore in-file macro definition with 0 line complaint for clang Tom de Vries
@ 2024-05-17 10:47 ` Tom de Vries
0 siblings, 0 replies; 2+ messages in thread
From: Tom de Vries @ 2024-05-17 10:47 UTC (permalink / raw)
To: gdb-patches
With test-case test.c:
...
#define NOARG
int main (void) { return 0; }
...
compiled with clang:
...
$ clang test.c -g3 -fdebug-macro
...
we get a complaint:
...
$ gdb -q -batch -iex "set complaints 5" a.out -ex "p main"
During symbol reading: macro debug info contains a malformed macro definition:
`NOARG'
...
and four others from pre-defined defs.
This is because the macro body is defined as "NOARG" instead of "NOARG ".
There's a comment in parse_macro_definition that describes why the space
should be there, quoting the standard.
My guess is that this is just a matter of interpretation. Either we have
- a name and empty definition, separated by a space, or
- a name and no definition, and therefore no space.
Anyway, this is long-standing and AFAICT intentional behaviour from clang, and
I think it's entirely reasonable, so drop the complaint for clang.
Likewise for '#define ARG(V)'.
Tested on x86_64-linux.
---
gdb/dwarf2/macro.c | 12 +++++++-----
gdb/testsuite/gdb.dwarf2/clang-cli-macro.exp | 4 ++++
2 files changed, 11 insertions(+), 5 deletions(-)
diff --git a/gdb/dwarf2/macro.c b/gdb/dwarf2/macro.c
index 66a40e88d3a..deaad155bef 100644
--- a/gdb/dwarf2/macro.c
+++ b/gdb/dwarf2/macro.c
@@ -103,7 +103,7 @@ consume_improper_spaces (const char *p, const char *body)
static void
parse_macro_definition (struct macro_source_file *file, int line,
- const char *body)
+ const char *body, struct dwarf2_cu *cu)
{
const char *p;
@@ -149,7 +149,8 @@ parse_macro_definition (struct macro_source_file *file, int line,
replacement = body + name_len + 1;
else
{
- dwarf2_macro_malformed_definition_complaint (body);
+ if (!producer_is_clang (cu))
+ dwarf2_macro_malformed_definition_complaint (body);
replacement = body + name_len;
}
@@ -213,7 +214,8 @@ parse_macro_definition (struct macro_source_file *file, int line,
else if (*p == '\0')
{
/* Complain, but do define it. */
- dwarf2_macro_malformed_definition_complaint (body);
+ if (!producer_is_clang (cu))
+ dwarf2_macro_malformed_definition_complaint (body);
macro_define_function (file, line, name.c_str (),
argc, (const char **) argv,
p);
@@ -569,7 +571,7 @@ dwarf_decode_macro_bytes (dwarf2_per_objfile *per_objfile,
line, current_file->filename);
}
else if (is_define)
- parse_macro_definition (current_file, line, body);
+ parse_macro_definition (current_file, line, body, cu);
else
{
gdb_assert (macinfo_type == DW_MACRO_undef
@@ -633,7 +635,7 @@ dwarf_decode_macro_bytes (dwarf2_per_objfile *per_objfile,
}
if (macinfo_type == DW_MACRO_define_strx)
- parse_macro_definition (current_file, line, body);
+ parse_macro_definition (current_file, line, body, cu);
else
macro_undef (current_file, line, body);
}
diff --git a/gdb/testsuite/gdb.dwarf2/clang-cli-macro.exp b/gdb/testsuite/gdb.dwarf2/clang-cli-macro.exp
index fdeaaff5afd..536132f6aa2 100644
--- a/gdb/testsuite/gdb.dwarf2/clang-cli-macro.exp
+++ b/gdb/testsuite/gdb.dwarf2/clang-cli-macro.exp
@@ -75,6 +75,8 @@ Dwarf::assemble $asm_file {
start_file 0 1
# A macro defined at line 1 of the main file.
define 1 "TWO 2"
+ define 2 "THREE"
+ define 3 "FOUR(ARG)"
end_file
define 0 "ONE 1"
}
@@ -97,3 +99,5 @@ if {![runto_main]} {
gdb_test "print TWO" "= 2" "print simple macro"
gdb_test "print ONE" "= 1" "print defined from CLI"
+gdb_test "info macro THREE" "\r\n#define THREE "
+gdb_test "info macro FOUR" "\r\n[string_to_regexp {#define FOUR(ARG) }]"
--
2.35.3
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2024-05-17 10:46 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-05-17 10:47 [PATCH 1/2] [gdb/macro] Ignore in-file macro definition with 0 line complaint for clang Tom de Vries
2024-05-17 10:47 ` [PATCH 2/2] [gdb/macro] Ignore malformed macro definition " Tom de Vries
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).