| Message ID | 20200628165752.5379-1-Nitika.Achra@amd.com |
|---|---|
| State | Superseded |
| Headers | show |
| Series |
|
| Related | show |
git tells me this when applying: Applying: DWARFv5: Handle DW_MACRO_define_strx and DW_MACRO_undef_strx macro entries. .git/rebase-apply/patch:50: trailing whitespace. warning: 1 line adds whitespace errors. please find and fix it. On 2020-06-28 12:57 p.m., nitachra wrote: > GDB is not able to print the macro values with -gdwarf-5 for clang > generated binaries. Clang is emitting DW_MACRO_define_strx and > DW_MACRO_undef_strx entries in .debug_macro section which are not > supported in GDB. This patch handles them. > > DW_MACRO_define_strx and DW_MACRO_undef_strx are added in DWARFv5. > They have two operands. The first operand encodes the line number of > the #define or #undef macro directive. The second operand identifies > a string; it is represented using an unsigned LEB128 encoded value, > which is interpreted as a zero-based index into an array of offsets > in the .debug_str_offsets section. This is as per the section 6.3.2.1 > of Dwarf Debugging Information Format Version 5. > > Test case used: > #define MAX_SIZE 10 > int main(void) > { > int size = 0; > size = size + MAX_SIZE; > > printf("\n The value of size is [%d]\n",size); > > return 0; > } > > clang -gdwarf-5 -fdebug-macro macro.c -o macro.out > > Before the patch: > > gdb -q macro.out -ex "start" -ex "p MAX_SIZE" > Reading symbols from macro.out... > Temporary breakpoint 1 at 0x4004df: file macro.c, line 7. > Starting program: /home/nitika/workspace/macro.out > > Temporary breakpoint 1, main () at macro.c:7 > 7 int size = 0; > No symbol "MAX_SIZE" in current context. > > After the patch: > > gdb -q macro.out -ex "start" -ex "p MAX_SIZE" > Reading symbols from macro.out... > Temporary breakpoint 1 at 0x4004df: file macro.c, line 7. > Starting program: /home/nitika/workspace/macro.out > > Temporary breakpoint 1, main () at macro.c:7 > 7 int size = 0; > $1 = 10 > > Tested by running the testsuite before and after the patch with > -gdwarf-5 and there is no increase in the number of test cases > that fails. Used clang 11.0.0. Hmm this is a bit misleading, as clang 11.0.0 is not released yet. You must have tested with a preview build. Please indicate the exact version. But thanks for specifying, because I only have clang 10 right now, and I didn't see DW_MACRO_define_strx getting generated. > > gdb/dwarf2/ChangeLog: There is no gdb/dwarf2/ChangeLog file. This will go in gdb/ChangeLog. > *macro.c (dwarf_decode_macro_bytes): Handle DW_MACRO_define_strx > and DW_MACRO_undef_strx. > (dwarf_decode_macros): Likewise Maybe it's just your mail client, but just to be sure, the format is: <tab>*<space><filename>... Wrapped lines are ligned with the *. > *read.c (dwarf_decode_macros): Pass str_offsets_base in the parameters > which is the value of DW_AT_str_offsets_base. > *macro.h (dwarf_decode_macros): Modify the definition to include > str_offsets_base. > > gdb/ChangeLog: > > *macrotab.c (macro_lookup_inclusion): In DWARF version 5 filename > includes the full directory path. > --- > gdb/dwarf2/macro.c | 63 +++++++++++++++++++++++++++++++++++++++++++--- > gdb/dwarf2/macro.h | 1 + > gdb/dwarf2/read.c | 4 ++- > gdb/macrotab.c | 3 ++- > 4 files changed, 66 insertions(+), 5 deletions(-) > > diff --git a/gdb/dwarf2/macro.c b/gdb/dwarf2/macro.c > index d047d806f8..2f5f3aa827 100644 > --- a/gdb/dwarf2/macro.c > +++ b/gdb/dwarf2/macro.c > @@ -427,6 +427,7 @@ dwarf_decode_macro_bytes (dwarf2_per_objfile *per_objfile, > const struct dwarf2_section_info *section, > int section_is_gnu, int section_is_dwz, > unsigned int offset_size, > + ULONGEST str_offsets_base, > htab_t include_hash) > { > struct objfile *objfile = per_objfile->objfile; > @@ -561,6 +562,50 @@ dwarf_decode_macro_bytes (dwarf2_per_objfile *per_objfile, > } > break; > > + case DW_MACRO_define_strx: > + case DW_MACRO_undef_strx: > + { > + unsigned int bytes_read; > + int line; > + int offset_index; > + const char* body; > + const gdb_byte* info_ptr; Space before asterisk. > + ULONGEST str_offset; Declare all of these on first use (where applicable, doesn't apply to `bytes_read` of course). > + > + line = read_unsigned_leb128 (abfd, mac_ptr, &bytes_read); > + mac_ptr += bytes_read; > + offset_index = read_unsigned_leb128 (abfd, mac_ptr, &bytes_read); > + mac_ptr += bytes_read; > + struct dwarf2_section_info str_offsets_section = per_objfile->per_bfd->str_offsets; This line is too long. When using a .dwo, the .debug_macro section is in the dwo (.debug_macro.dwo)? If so, should we use the string offset and string sections from the dwo? We don't have a testsuite board for DWARF5 + split-dwarf currently, I think it would be useful. > + str_offsets_section.read (objfile); > + info_ptr = (str_offsets_section.buffer > + + str_offsets_base > + + offset_index * offset_size); > + if (offset_size == 4) > + str_offset = bfd_get_32 (abfd, info_ptr); > + else > + str_offset = bfd_get_64 (abfd, info_ptr); Can we get some bound checking here, to avoid reading past the section buffer? > + body = per_objfile->per_bfd->str.read_string (objfile, > + str_offset, > + "DW_FORM_strx"); The "DW_FORM_strx" is used for error reporting inside `read_string`. I'm not sure it's exact DW_FORM_strx here, because forms are used for attribute values, which is not the case here. Perhaps you could use "DW_MACRO_define_strx" and "DW_MACRO_undef_strx", it would produce error messages like: DW_MACRO_define_strx used without %s section [in module %s] DW_MACRO_define_strx pointing outside of %s section [in module %s] which I think make sense. > + > + if (! current_file) current_file == nullptr > + { > + /* DWARF violation as no main source is present. */ > + complaint (_("debug info with no main source gives macro %s " > + "on line %d: %s"), > + macinfo_type == DW_MACRO_define_strx ? _("definition") > + : _("undefinition"), line, body); > + break; There are spaces that should be tabs here. Simon
>> + struct dwarf2_section_info str_offsets_section = per_objfile->per_bfd->str_offsets;
Simon> This line is too long.
It's also slightly bad to copy a section like this. In weird cases I
think it can lead to it being read more than once.
We should probably delete operator= here, but that's a pain right now.
I think I have some patches toward this somewhere, but I haven't
finished them yet. (Also, I think the DWP code does do this at the
moment.)
Better here to use a reference or a pointer.
Tom
On 2020-07-11 6:54 p.m., Tom Tromey wrote: >>> + struct dwarf2_section_info str_offsets_section = per_objfile->per_bfd->str_offsets; > > Simon> This line is too long. > > It's also slightly bad to copy a section like this. In weird cases I > think it can lead to it being read more than once. > > We should probably delete operator= here, but that's a pain right now. > I think I have some patches toward this somewhere, but I haven't > finished them yet. (Also, I think the DWP code does do this at the > moment.) > > Better here to use a reference or a pointer. > > Tom > Ah, I didn't notice that it wasn't a pointer, thanks. Simon
diff --git a/gdb/dwarf2/macro.c b/gdb/dwarf2/macro.c index d047d806f8..2f5f3aa827 100644 --- a/gdb/dwarf2/macro.c +++ b/gdb/dwarf2/macro.c @@ -427,6 +427,7 @@ dwarf_decode_macro_bytes (dwarf2_per_objfile *per_objfile, const struct dwarf2_section_info *section, int section_is_gnu, int section_is_dwz, unsigned int offset_size, + ULONGEST str_offsets_base, htab_t include_hash) { struct objfile *objfile = per_objfile->objfile; @@ -561,6 +562,50 @@ dwarf_decode_macro_bytes (dwarf2_per_objfile *per_objfile, } break; + case DW_MACRO_define_strx: + case DW_MACRO_undef_strx: + { + unsigned int bytes_read; + int line; + int offset_index; + const char* body; + const gdb_byte* info_ptr; + ULONGEST str_offset; + + line = read_unsigned_leb128 (abfd, mac_ptr, &bytes_read); + mac_ptr += bytes_read; + offset_index = read_unsigned_leb128 (abfd, mac_ptr, &bytes_read); + mac_ptr += bytes_read; + struct dwarf2_section_info str_offsets_section = per_objfile->per_bfd->str_offsets; + str_offsets_section.read (objfile); + info_ptr = (str_offsets_section.buffer + + str_offsets_base + + offset_index * offset_size); + if (offset_size == 4) + str_offset = bfd_get_32 (abfd, info_ptr); + else + str_offset = bfd_get_64 (abfd, info_ptr); + body = per_objfile->per_bfd->str.read_string (objfile, + str_offset, + "DW_FORM_strx"); + + if (! current_file) + { + /* DWARF violation as no main source is present. */ + complaint (_("debug info with no main source gives macro %s " + "on line %d: %s"), + macinfo_type == DW_MACRO_define_strx ? _("definition") + : _("undefinition"), line, body); + break; + } + + if (macinfo_type == DW_MACRO_define_strx) + parse_macro_definition (current_file, line, body); + else + macro_undef (current_file, line, body); + } + break; + case DW_MACRO_start_file: { unsigned int bytes_read; @@ -671,7 +716,7 @@ dwarf_decode_macro_bytes (dwarf2_per_objfile *per_objfile, new_mac_ptr, include_mac_end, current_file, lh, section, section_is_gnu, is_dwz, offset_size, - include_hash); + str_offsets_base, include_hash); htab_remove_elt (include_hash, (void *) new_mac_ptr); } @@ -712,7 +757,8 @@ dwarf_decode_macros (dwarf2_per_objfile *per_objfile, buildsym_compunit *builder, const dwarf2_section_info *section, const struct line_header *lh, unsigned int offset_size, - unsigned int offset, int section_is_gnu) + unsigned int offset, ULONGEST str_offsets_base, + int section_is_gnu) { bfd *abfd; const gdb_byte *mac_ptr, *mac_end; @@ -814,6 +860,17 @@ dwarf_decode_macros (dwarf2_per_objfile *per_objfile, mac_ptr += offset_size; } break; + case DW_MACRO_define_strx: + case DW_MACRO_undef_strx: + { + unsigned int bytes_read; + + read_unsigned_leb128 (abfd, mac_ptr, &bytes_read); + mac_ptr += bytes_read; + read_unsigned_leb128 (abfd, mac_ptr, &bytes_read); + mac_ptr += bytes_read; + } + break; case DW_MACRO_import: case DW_MACRO_import_sup: @@ -861,5 +918,5 @@ dwarf_decode_macros (dwarf2_per_objfile *per_objfile, *slot = (void *) mac_ptr; dwarf_decode_macro_bytes (per_objfile, builder, abfd, mac_ptr, mac_end, current_file, lh, section, section_is_gnu, 0, - offset_size, include_hash.get ()); + offset_size, str_offsets_base, include_hash.get ()); } diff --git a/gdb/dwarf2/macro.h b/gdb/dwarf2/macro.h index d874068947..dc396e1b70 100644 --- a/gdb/dwarf2/macro.h +++ b/gdb/dwarf2/macro.h @@ -28,6 +28,7 @@ extern void dwarf_decode_macros (dwarf2_per_objfile *per_objfile, const struct line_header *lh, unsigned int offset_size, unsigned int offset, + ULONGEST str_offsets_base, int section_is_gnu); #endif /* GDB_DWARF2_MACRO_H */ diff --git a/gdb/dwarf2/read.c b/gdb/dwarf2/read.c index e3073fe43c..a0e197c748 100644 --- a/gdb/dwarf2/read.c +++ b/gdb/dwarf2/read.c @@ -23353,8 +23353,10 @@ dwarf_decode_macros (struct dwarf2_cu *cu, unsigned int offset, buildsym_compunit *builder = cu->get_builder (); + ULONGEST str_offsets_base = *cu->str_offsets_base; + dwarf_decode_macros (per_objfile, builder, section, lh, - offset_size, offset, section_is_gnu); + offset_size, offset, str_offsets_base, section_is_gnu); } /* Return the .debug_loc section to use for CU. diff --git a/gdb/macrotab.c b/gdb/macrotab.c index 63cd30148a..606f084bc9 100644 --- a/gdb/macrotab.c +++ b/gdb/macrotab.c @@ -507,7 +507,8 @@ struct macro_source_file * macro_lookup_inclusion (struct macro_source_file *source, const char *name) { /* Is SOURCE itself named NAME? */ - if (filename_cmp (name, source->filename) == 0) + if (filename_cmp (name, source->filename) == 0 || + filename_cmp (basename (name), basename (source->filename)) == 0) return source; /* It's not us. Try all our children, and return the lowest. */
GDB is not able to print the macro values with -gdwarf-5 for clang generated binaries. Clang is emitting DW_MACRO_define_strx and DW_MACRO_undef_strx entries in .debug_macro section which are not supported in GDB. This patch handles them. DW_MACRO_define_strx and DW_MACRO_undef_strx are added in DWARFv5. They have two operands. The first operand encodes the line number of the #define or #undef macro directive. The second operand identifies a string; it is represented using an unsigned LEB128 encoded value, which is interpreted as a zero-based index into an array of offsets in the .debug_str_offsets section. This is as per the section 6.3.2.1 of Dwarf Debugging Information Format Version 5. Test case used: #define MAX_SIZE 10 int main(void) { int size = 0; size = size + MAX_SIZE; printf("\n The value of size is [%d]\n",size); return 0; } clang -gdwarf-5 -fdebug-macro macro.c -o macro.out Before the patch: gdb -q macro.out -ex "start" -ex "p MAX_SIZE" Reading symbols from macro.out... Temporary breakpoint 1 at 0x4004df: file macro.c, line 7. Starting program: /home/nitika/workspace/macro.out Temporary breakpoint 1, main () at macro.c:7 7 int size = 0; No symbol "MAX_SIZE" in current context. After the patch: gdb -q macro.out -ex "start" -ex "p MAX_SIZE" Reading symbols from macro.out... Temporary breakpoint 1 at 0x4004df: file macro.c, line 7. Starting program: /home/nitika/workspace/macro.out Temporary breakpoint 1, main () at macro.c:7 7 int size = 0; $1 = 10 Tested by running the testsuite before and after the patch with -gdwarf-5 and there is no increase in the number of test cases that fails. Used clang 11.0.0. gdb/dwarf2/ChangeLog: *macro.c (dwarf_decode_macro_bytes): Handle DW_MACRO_define_strx and DW_MACRO_undef_strx. (dwarf_decode_macros): Likewise *read.c (dwarf_decode_macros): Pass str_offsets_base in the parameters which is the value of DW_AT_str_offsets_base. *macro.h (dwarf_decode_macros): Modify the definition to include str_offsets_base. gdb/ChangeLog: *macrotab.c (macro_lookup_inclusion): In DWARF version 5 filename includes the full directory path. --- gdb/dwarf2/macro.c | 63 +++++++++++++++++++++++++++++++++++++++++++--- gdb/dwarf2/macro.h | 1 + gdb/dwarf2/read.c | 4 ++- gdb/macrotab.c | 3 ++- 4 files changed, 66 insertions(+), 5 deletions(-) -- 2.17.1