[Bug symtab/28221] New: [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[Bug symtab/28221] New: [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

            Bug ID: 28221
           Summary: [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info
                    line func
           Product: gdb
           Version: HEAD
            Status: NEW
          Severity: normal
          Priority: P2
         Component: symtab
          Assignee: unassigned at sourceware dot org
          Reporter: vries at gcc dot gnu.org
  Target Milestone: ---

With target board readnow/-fno-PIE/-no-pie [on ubuntu 18.04, we use the -no-pie
to work around PR12497 ] we run into:
...
(gdb) PASS: gdb.dwarf2/dw2-ranges.exp: info line main
info line func^M
No line number information available for address 0x400528 <func>^M
(gdb) FAIL: gdb.dwarf2/dw2-ranges.exp: info line func
...

The problem is that the while the code from dw2-ranges3.c:
...
00000000004004a9 <main3>:
  4004a9:       55                      push   %rbp
  4004aa:       48 89 e5                mov    %rsp,%rbp
  4004ad:       90                      nop
  4004ae:       5d                      pop    %rbp
  4004af:       c3                      retq
...
is in the 4004a9-4004af range, the blockvector ends at 0x40053b:
...
$ gdb -q -batch outputs/gdb.dwarf2/dw2-ranges/dw2-ranges -ex start -ex "maint
expand-symtabs" -ex "maint print symbols"
  ...
Symtab for file
/home/vries/gdb/build/gdb/testsuite/outputs/gdb.dwarf2/dw2-ranges/dw2-ranges3.s
at 0x55b2b81908b0
Read from object file
/home/vries/gdb/build/gdb/testsuite/outputs/gdb.dwarf2/dw2-ranges/dw2-ranges
(0x55b2b81374b0)
Language: asm

Line table:

 line 8 at 0x4004a9     (stmt)
 line 11 at 0x4004aa    (stmt)
 line 13 at 0x4004ad    (stmt)
 line 14 at 0x4004ae    (stmt)
 line 16 at 0x4004af    (stmt)

Blockvector:

block #000, object at 0x55b2b8190830, 0 syms/buckets in 0x4004a9..0x40053b
  block #001, object at 0x55b2b81907d0 under 0x55b2b8190830, 0 syms/buckets in
0x4004a9..0x40053b
...

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug symtab/28221] [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

--- Comment #1 from Tom de Vries <vries at gcc dot gnu.org> ---
Hack that makes the test pass:
...
diff --git a/gdb/dbxread.c b/gdb/dbxread.c
index cf35880b509..7fd570cf515 100644
--- a/gdb/dbxread.c
+++ b/gdb/dbxread.c
@@ -1896,10 +1896,13 @@ read_dbx_symtab (minimal_symbol_reader &reader,
         : lowest_text_address)
        + text_size;

+      if (pst->raw_text_high() == pst->raw_text_low ())
+       pst->set_text_high (pst->raw_text_low () + 1);
+
       dbx_end_psymtab (objfile, partial_symtabs,
                       pst, psymtab_include_list, includes_used,
                       symnum * symbol_size,
-                      (text_end > pst->raw_text_high ()
+                      (text_end < pst->raw_text_high ()
                        ? text_end : pst->raw_text_high ()),
                       dependency_list, dependencies_used, textlow_not_set);
     }
...

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug symtab/28221] [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

--- Comment #2 from Tom de Vries <vries at gcc dot gnu.org> ---
Passes with:
...
diff --git a/gdb/testsuite/gdb.dwarf2/dw2-ranges.exp
b/gdb/testsuite/gdb.dwarf2/dw2-ranges.exp
index 8566e0034cf..2da3308e661 100644
--- a/gdb/testsuite/gdb.dwarf2/dw2-ranges.exp
+++ b/gdb/testsuite/gdb.dwarf2/dw2-ranges.exp
@@ -52,11 +52,7 @@ if {[gdb_compile "${srcdir}/${subdir}/${srcfile2}"
"${asmfile2}" assembly {}] !=
     return -1
 }

-# STABS compilation is intentional as it tests the STABS lookup where is no
-# partial_symtab->psymtabs_addrmap supported for that CU (Compilation Unit) in
-# a file containing psymtabs_addrmap-supporting DWARF CUs.
-if {[gdb_compile "${srcdir}/${subdir}/${srcfile3}" "${asmfile3}" assembly {}]
!= ""
-    || [gdb_compile "${asmfile3}" "${objfile3}" object
{additional_flags=-gstabs}] != "" } {
+if {[gdb_compile "${srcdir}/${subdir}/${srcfile3}" "${objfile3}" object
{additional_flags=-gstabs}] != "" } {
     return -1
 }

...

So this seems again an issue with gas-generated stabs.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug symtab/28221] [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

--- Comment #3 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #2)
> So this seems again an issue with gas-generated stabs.

Indeed, the compiler-generated stabs look like:
...
28     SLINE  0      19     0000000000000000 0      
29     SLINE  0      20     0000000000000004 0      
30     SO     0      0      00000000004004b0 0      
...
so they contain an end address.

Looking at the assembly file, that's generated from:
...
        .stabs  "",100,0,0,.Letext0
.Letext0:
...

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug symtab/28221] [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

--- Comment #4 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #3)
> Indeed, the compiler-generated stabs look like:

And the gas-generated stabs:
...
Contents of .stab section:

Symnum n_type n_othr n_desc n_value  n_strx String

-1     HdrSym 0      6      0000000000000051 1     
0      SO     0      0      00000000004004a9 1     
/home/vries/gdb/build/gdb/testsuite/outputs/gdb.dwarf2/dw2-ranges/dw2-ranges3.s
1      SLINE  0      8      00000000004004a9 0      
2      SLINE  0      11     00000000004004aa 0      
3      SLINE  0      13     00000000004004ad 0      
4      SLINE  0      14     00000000004004ae 0      
5      SLINE  0      16     00000000004004af 0      
...
ends with SLINE instead, so the end-of-text-section is used as CU end address.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug symtab/28221] [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

--- Comment #5 from Tom de Vries <vries at gcc dot gnu.org> ---
With this patch:
...
diff --git a/gdb/testsuite/gdb.dwarf2/dw2-ranges.exp
b/gdb/testsuite/gdb.dwarf2/dw2-ranges.exp
index 8566e0034cf..360e3590e29 100644
--- a/gdb/testsuite/gdb.dwarf2/dw2-ranges.exp
+++ b/gdb/testsuite/gdb.dwarf2/dw2-ranges.exp
@@ -36,6 +36,7 @@ standard_testfile .c -2.c -3.c
 set asmfile [standard_output_file ${testfile}.s]
 set asmfile2 [standard_output_file ${testfile}2.s]
 set asmfile3 [standard_output_file ${testfile}3.s]
+set asmfiletmp [standard_output_file ${testfile}tmp.s]
 set objfile [standard_output_file ${testfile}.o]
 set objfile2 [standard_output_file ${testfile}2.o]
 set objfile3 [standard_output_file ${testfile}3.o]
@@ -55,8 +56,16 @@ if {[gdb_compile "${srcdir}/${subdir}/${srcfile2}"
"${asmfile2}" assembly {}] !=
 # STABS compilation is intentional as it tests the STABS lookup where is no
 # partial_symtab->psymtabs_addrmap supported for that CU (Compilation Unit) in
 # a file containing psymtabs_addrmap-supporting DWARF CUs.
-if {[gdb_compile "${srcdir}/${subdir}/${srcfile3}" "${asmfile3}" assembly {}]
!= ""
-    || [gdb_compile "${asmfile3}" "${objfile3}" object
{additional_flags=-gstabs}] != "" } {
+if { [gdb_compile "${srcdir}/${subdir}/${srcfile3}" "${asmfiletmp}" assembly
{}] != "" } {
+    return -1
+}
+
+exec echo ".func main3" > $asmfile3
+exec sed -i "s/.*GNU-stack.*//" $asmfiletmp
+exec cat $asmfiletmp >> $asmfile3
+exec echo ".endfunc" >> $asmfile3
+
+if { [gdb_compile "${asmfile3}" "${objfile3}" object
{additional_flags=-gstabs}] != "" } {
     return -1
 }

...
I get:
...
.func main3
        .file   "dw2-ranges-3.c"
        .text
        .globl  main3
        .type   main3, @function
main3:
.LFB0:
        .cfi_startproc
        pushq   %rbp
        .cfi_def_cfa_offset 16
        .cfi_offset 6, -16
        movq    %rsp, %rbp
        .cfi_def_cfa_register 6
        nop
        popq    %rbp
        .cfi_def_cfa 7, 8
        ret
        .cfi_endproc
.LFE0:
        .size   main3, .-main3
        .ident  "GCC: (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0"

.endfunc
...
and:
...
Contents of .stab section:

Symnum n_type n_othr n_desc n_value  n_strx String

-1     HdrSym 0      9      0000000000000064 1     
0      SO     0      0      00000000004004a9 1     
/home/vries/gdb/build/gdb/testsuite/outputs/gdb.dwarf2/dw2-ranges/dw2-ranges3.s
1      LSYM   0      0      0000000000000000 81     void:t1=1
2      FUN    0      2      00000000004004a9 91     main3:F1
3      SLINE  0      9      0000000000000000 0      
4      SLINE  0      12     0000000000000001 0      
5      SLINE  0      14     0000000000000004 0      
6      SLINE  0      15     0000000000000005 0      
7      SLINE  0      17     0000000000000006 0      
8      FUN    0      0      0000000000000007 0      
...
and the test still fails.

Possibly because the terminating FUN uses relative address?  The
compiler-generated version uses absolute address.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug symtab/28221] [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

--- Comment #6 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #5)
> Possibly because the terminating FUN uses relative address?  The
> compiler-generated version uses absolute address.

I'd like to try:
...
diff --git a/gas/stabs.c b/gas/stabs.c
index 8da1147c561..c9e8b886ef9 100644
--- a/gas/stabs.c
+++ b/gas/stabs.c
@@ -692,7 +692,7 @@ stabs_generate_asm_endfunc (const char *funcname
ATTRIBUTE_UNUSED,
   ++label_count;
   colon (sym);

-  if (asprintf (&buf, "\"\",%d,0,0,%s-%s", N_FUN, sym, startlabname) == -1)
+  if (asprintf (&buf, "\"\",%d,0,0,%s", N_FUN, sym) == -1)
     as_fatal ("%s", xstrerror (errno));

   temp_ilp (buf);
...
but on my ubuntu 18.04.5 setup I can't use current gas with the gdb test-case,
because I run into:
...
gdb compile failed, /usr/bin/ld: warning:
/home/vries/gdb/build/gdb/testsuite/outputs/gdb.dwarf2/dw2-ranges/dw2-ranges.o:
unsupported GNU_PROPERTY_TYPE (5) type: 0xc0010002
...

Maybe a question of ubuntu/debian-specific patches, maybe the version is too
new.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug symtab/28221] [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

--- Comment #7 from Tom de Vries <vries at gcc dot gnu.org> ---
Tentative gas patch, fixes PR:
...
diff --git a/gas/read.c b/gas/read.c
index 7bf52f11218..d75815fb99a 100644
--- a/gas/read.c
+++ b/gas/read.c
@@ -248,6 +248,7 @@ static segT get_known_segmented_expression (expressionS *
expP);
 static void pobegin (void);
 static size_t get_non_macro_line_sb (sb *);
 static void generate_file_debug (void);
+static void generate_file_debug_end (void);
 static char *_find_end_of_line (char *, int, int, int);
 ^L
 void
@@ -1363,6 +1364,8 @@ read_a_source_file (const char *name)
     }

  quit:
+  generate_file_debug_end ();
+
   symbol_set_value_now (&dot_symbol);

 #ifdef HANDLE_BUNDLE
@@ -6066,6 +6069,13 @@ generate_file_debug (void)
     stabs_generate_asm_file ();
 }

+static void
+generate_file_debug_end (void)
+{
+  if (debug_type == DEBUG_STABS)
+    stabs_generate_asm_file_end ();
+}
+
 /* Output line number debugging information for the current source line.  */

 void
diff --git a/gas/read.h b/gas/read.h
index eebdc4e7452..651f0701bf0 100644
--- a/gas/read.h
+++ b/gas/read.h
@@ -144,6 +144,7 @@ extern void read_print_statistics (FILE *);
 extern char *read_symbol_name (void);
 extern unsigned sizeof_leb128 (valueT, int);
 extern void stabs_generate_asm_file (void);
+extern void stabs_generate_asm_file_end (void);
 extern void stabs_generate_asm_lineno (void);
 extern void stabs_generate_asm_func (const char *, const char *);
 extern void stabs_generate_asm_endfunc (const char *, const char *);
diff --git a/gas/stabs.c b/gas/stabs.c
index 73d1361423c..946831d9c18 100644
--- a/gas/stabs.c
+++ b/gas/stabs.c
@@ -502,6 +502,13 @@ stabs_generate_asm_file (void)
   generate_asm_file (N_SO, file);
 }

+void
+stabs_generate_asm_file_end (void)
+{
+  subseg_set (text_section, 0);
+  generate_asm_file (N_SO, "");
+}
+
 /* Generate stabs debugging information to denote the source file.
    TYPE is one of N_SO, N_SOL.  */

...

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug symtab/28221] [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

--- Comment #8 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #7)
> Tentative gas patch, fixes PR:

With which we have (with .exp modification to add .func prefix):
...
$ objdump -G outputs/gdb.dwarf2/dw2-ranges/dw2-ranges

outputs/gdb.dwarf2/dw2-ranges/dw2-ranges:     file format elf64-x86-64

Contents of .stab section:

Symnum n_type n_othr n_desc n_value  n_strx String

-1     HdrSym 0      9      0000000000000064 1     
0      SO     0      0      00000000004004a9 1     
/home/vries/gdb/build/gdb/testsuite/outputs/gdb.dwarf2/dw2-ranges/dw2-ranges3.s
1      LSYM   0      0      0000000000000000 81     void:t1=1
2      FUN    0      2      00000000004004a9 91     main3:F1
3      SLINE  0      9      0000000000000000 0      
4      SLINE  0      12     0000000000000001 0      
5      SLINE  0      14     0000000000000004 0      
6      SLINE  0      15     0000000000000005 0      
7      SLINE  0      17     0000000000000006 0      
8      SO     0      0      00000000004004b0 0      
...

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug symtab/28221] [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

--- Comment #9 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #7)
> Tentative gas patch, fixes PR:

Filed at PR28233 - "[gas, --gstabs] Generate stabs more similar to gcc"

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug symtab/28221] [readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[cvs-commit at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

--- Comment #10 from cvs-commit at gcc dot gnu.org <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Tom de Vries <vries@sourceware.org>:

https://sourceware.org/git/gitweb.cgi?p=binutils-gdb.git;h=74cdc3e0888f1ffc3973a4826e85dd32e5a51037

commit 74cdc3e0888f1ffc3973a4826e85dd32e5a51037
Author: Tom de Vries <tdevries@suse.de>
Date:   Mon Aug 23 12:08:25 2021 +0200

    [gdb/testsuite] Use compiler-generated instead of gas-generated stabs

    The test-case gdb.dwarf2/dw2-ranges.exp is the only one in the gdb
testsuite
    that uses gas-generated stabs.

    While the use seems natural alongside the use of gas-generated dwarf in the
    same test-case, there are a few known issues, filed on the gdb side as:
    - PR symtab/12497 - "stabs: PIE relocation does not work"
    - PR symtab/28221 - "[readnow, stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: \
      info line func"
    and on the gas side as:
    - PR gas/28233 - "[gas, --gstabs] Generate stabs more similar to gcc"

    The test-case contains a KFAIL for PR12497, but it's outdated and fails to
    trigger.

    The intention of the test-case is to test gas-generated dwarf, and using
    gcc-generated stabs instead of gas-generated stabs works fine.

    Supporting compiler-generated stabs is already a corner-case for gdb, and
    there's no current commitment/incentive to support/workaround gas-generated
    stabs, which can be considered a corner-case of a corner-case.

    Work around these problem by using compiler-generated stabs in the
test-case.

    Tested on x86_64-linux.

    gdb/testsuite/ChangeLog:

    2021-08-22  Tom de Vries  <tdevries@suse.de>

            * gdb.dwarf2/dw2-ranges.exp: Use compiler-generated stabs.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug symtab/28221] [readnow, gas-stabs] FAIL: gdb.dwarf2/dw2-ranges.exp: info line func

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=28221

Tom de Vries <vries at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |WONTFIX
            Summary|[readnow, stabs] FAIL:      |[readnow, gas-stabs] FAIL:
                   |gdb.dwarf2/dw2-ranges.exp:  |gdb.dwarf2/dw2-ranges.exp:
                   |info line func              |info line func
             Status|NEW                         |RESOLVED

--- Comment #11 from Tom de Vries <vries at gcc dot gnu.org> ---
The test-case has been updated to not use gas-generated stabs anymore.

I'm marking this PR resolved-wontfix.

If there's follow-up on PR gas/28233, we can re-introduce gas-generates stabs
to the test-case, using an axis stabs_method=[list compiler assembler].

-- 
You are receiving this mail because:
You are on the CC list for the bug.