[PING*2] gdb/linespec.c: Fix missing source file during breakpoint re-set

Message ID 20230106235607.498218-1-amerey@redhat.com
State Committed
Commit 7dd38e31d67c2548b52bea313ab18e40824c05da
Headers
Series [PING*2] gdb/linespec.c: Fix missing source file during breakpoint re-set |

Commit Message

Aaron Merey Jan. 6, 2023, 11:56 p.m. UTC
  Ping. I tweaked the testcase so that it applies to the master branch.

Thanks,
Aaron

---
During breakpoint re-setting, the source_filename of an
explicit_location_spec is used to lookup the symtabs associated with
the breakpoint being re-set.  This source_filename is compared with each
known symtab filename in order to retrieve the breakpoint's symtabs.

However the source_filename may have been originally copied from a
symtab's fullname (the path where GDB found the source file) when the
breakpoint was first created.  If a breakpoint symtab's filename and
fullname differ and there is no substitute-path rule that converts the
fullname to the filename, this will cause a NOT_FOUND_ERROR to be thrown
during re-setting.

Fix this by using a symtab's filename to set the explicit_location_spec
source_filename instead of the symtab's fullname.
---
 gdb/linespec.c                                         | 4 ++--
 gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp | 9 +++++++++
 2 files changed, 11 insertions(+), 2 deletions(-)
  

Comments

Tom Tromey Jan. 9, 2023, 7:27 p.m. UTC | #1
>>>>> "Aaron" == Aaron Merey via Gdb-patches <gdb-patches@sourceware.org> writes:

Aaron> Ping. I tweaked the testcase so that it applies to the master branch.

Oops, I reviewed the wrong version.
This one is ok.  Thank you.

Tom
  
Aaron Merey Jan. 10, 2023, 1:15 a.m. UTC | #2
On Mon, Jan 9, 2023 at 2:28 PM Tom Tromey <tom@tromey.com> wrote:
> Oops, I reviewed the wrong version.
> This one is ok.  Thank you.

Thanks Tom, pushed as commit 7dd38e31d67.

On Mon, Jan 9, 2023 at 2:27 PM Tom Tromey <tom@tromey.com> wrote:
> It's pretty hard to review a patch like this, since it's hard to know
> what impact it might have.  However I am mostly fine with it (see the
> end), assuming you regression-tested it.

Regression tested on x86_64 F36 and F37.

Aaron
  
Andrew Burgess Feb. 1, 2023, 5:37 p.m. UTC | #3
Aaron Merey via Gdb-patches <gdb-patches@sourceware.org> writes:

> Ping. I tweaked the testcase so that it applies to the master branch.
>
> Thanks,
> Aaron
>
> ---
> During breakpoint re-setting, the source_filename of an
> explicit_location_spec is used to lookup the symtabs associated with
> the breakpoint being re-set.  This source_filename is compared with each
> known symtab filename in order to retrieve the breakpoint's symtabs.
>
> However the source_filename may have been originally copied from a
> symtab's fullname (the path where GDB found the source file) when the
> breakpoint was first created.  If a breakpoint symtab's filename and
> fullname differ and there is no substitute-path rule that converts the
> fullname to the filename, this will cause a NOT_FOUND_ERROR to be thrown
> during re-setting.
>
> Fix this by using a symtab's filename to set the explicit_location_spec
> source_filename instead of the symtab's fullname.
> ---
>  gdb/linespec.c                                         | 4 ++--
>  gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp | 9 +++++++++
>  2 files changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/gdb/linespec.c b/gdb/linespec.c
> index e9339c3338c..6db0f02e318 100644
> --- a/gdb/linespec.c
> +++ b/gdb/linespec.c
> @@ -2283,13 +2283,13 @@ convert_linespec_to_sals (struct linespec_state *state, linespec *ls)
>  	/* Make sure we have a filename for canonicalization.  */
>  	if (ls->explicit_loc.source_filename == NULL)
>  	  {
> -	    const char *fullname = symtab_to_fullname (state->default_symtab);
> +	    const char *filename = state->default_symtab->filename;
>  
>  	    /* It may be more appropriate to keep DEFAULT_SYMTAB in its symtab
>  	       form so that displaying SOURCE_FILENAME can follow the current
>  	       FILENAME_DISPLAY_STRING setting.  But as it is used only rarely
>  	       it has been kept for code simplicity only in absolute form.  */
> -	    ls->explicit_loc.source_filename = xstrdup (fullname);
> +	    ls->explicit_loc.source_filename = xstrdup (filename);
>  	  }
>      }
>    else
> diff --git a/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp b/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp
> index 6d5af1370b0..bfe6e639e4f 100644
> --- a/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp
> +++ b/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp
> @@ -224,6 +224,15 @@ proc_with_prefix local_url { } {
>      set lineno [gdb_get_line_number "Breakpoint here"]
>      gdb_test "list $lineno" "return 0;\[^\r\n\]+Breakpoint here\\. .*"
>  
> +    # Verify that a breakpoint re-sets correctly when the actual location
> +    # of the source file in the debuginfod client cache differs from
> +    # the contents of DW_AT_comp_dir and DW_AT_name.
> +    gdb_test "set cwd $debugdir" "" "file [file tail $binfile] cwd"
> +    gdb_test_no_output "del breakpoint 1"
> +    gdb_test "break $lineno" "Breakpoint 2 at.*file.*"
> +    gdb_test "run" "Breakpoint 2.*" \
> +	"file [file tail $binfile] set breakpoint"

Calling "run" like this will not work when testing with the
native-gdbserver board, which can be done like:

  make check-gdb RUNTESTFLAGS="--target_board=native-gdbserver"

The other board that is worth testing with is native-extended-gdbserver.

I've pushed the patch below which fixes the test to work with the
native-gdbserver board.  I checked that the test still fails if I revert
the GDB fix in your above patch.

Thanks,
Andrew

---

commit cded17bfca35566fa4d36e9ec06fa071bd7dab17
Author: Andrew Burgess <aburgess@redhat.com>
Date:   Wed Feb 1 17:09:47 2023 +0000

    gdb/testsuite: fix fetch_src_and_symbols.exp with native-gdbserver board
    
    I noticed that the gdb.debuginfod/fetch_src_and_symbols.exp script
    doesn't work with the native-gdbserver board, I see this error:
    
      ERROR: tcl error sourcing /tmp/build/gdb/testsuite/../../../src/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp.
      ERROR: gdbserver does not support run without extended-remote
          while executing
      "error "gdbserver does not support $command without extended-remote""
          (procedure "gdb_test_multiple" line 51)
          invoked from within
    
    This was introduced with this commit:
    
      commit 7dd38e31d67c2548b52bea313ab18e40824c05da
      Date:   Fri Jan 6 18:45:27 2023 -0500
    
          gdb/linespec.c: Fix missing source file during breakpoint re-set
    
    The problem is that the above commit introduces a direct use of the
    "run" command, which doesn't work with 'target remote' targets, as
    exercised by the native-gdbserver board.
    
    To avoid this, in this commit I switch to using runto_main.  However,
    calling runto_main will, by default, delete all the currently set
    breakpoints.  As the point of the above commit was to check that a
    breakpoint set before stating an inferior would be correctly re-set,
    we need to avoid this breakpoint deleting behaviour.
    
    To do this I make use of with_override, and override the
    delete_breakpoints proc with a dummy proc which does nothing.
    
    By reverting the GDB changes in commit 7dd38e31d67c I have confirmed
    that even after my changes in this commit, the test still fails.  But
    with the fixes in commit 7dd38e31d67c, this test now passed using the
    unix, native-gdbserver, and native-extended-gdbserver boards.

diff --git a/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp b/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp
index c9cd8a30a1c..8158c5c3cc6 100644
--- a/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp
+++ b/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp
@@ -195,6 +195,12 @@ proc test_urls {urls pattern_re test} {
 	$test
 }
 
+# Used as a replacement for delete_breakpoints while calling
+# runto_main in one case where we don't want to delete all the
+# breakpoints.
+proc disable_delete_breakpoints {} {
+}
+
 # Uses the global variables DEBUGDIR and DB which are setup elsewhere
 # in this script.
 #
@@ -214,13 +220,12 @@ proc_with_prefix local_url { } {
 
     # GDB should now find the symbol and source files.
     clean_restart
-    set enable_debuginfod_question \
-	"Enable debuginfod for this session. \\(y or \\\[n\\\]\\) "
-    gdb_test "file $binfile" "" "file [file tail $binfile]" \
-	$enable_debuginfod_question "y"
+    gdb_test_no_output "set debuginfod enabled on" \
+	"enabled debuginfod for initial test"
+    gdb_load $binfile
     gdb_test_no_output "set substitute-path $outputdir /dev/null" \
 	"set substitute-path"
-    gdb_test "br main" "Breakpoint 1 at.*file.*"
+
     set lineno [gdb_get_line_number "Breakpoint here"]
     gdb_test "list $lineno" "return 0;\[^\r\n\]+Breakpoint here\\. .*"
 
@@ -228,12 +233,18 @@ proc_with_prefix local_url { } {
     # of the source file in the debuginfod client cache differs from
     # the contents of DW_AT_comp_dir and DW_AT_name.
     gdb_test "set cwd $debugdir" "" "file [file tail $binfile] cwd"
-    gdb_test_no_output "del breakpoint 1"
-    gdb_test "break $lineno" "Breakpoint 2 at.*file.*"
-    gdb_test "run" "Breakpoint 2.*" \
-	"file [file tail $binfile] set breakpoint"
+    gdb_breakpoint $lineno
+    with_override delete_breakpoints disable_delete_breakpoints {
+	if {![runto_main]} {
+	    return
+	}
+	gdb_continue_to_breakpoint "runto breakpoint in main" \
+	    ".* Breakpoint here\\. .*"
+    }
 
     # GDB should now find the executable file.
+    set enable_debuginfod_question \
+	"Enable debuginfod for this session. \\(y or \\\[n\\\]\\) "
     clean_restart
     gdb_test "core $::corefile" ".*return 0.*" "file [file tail $::corefile]" \
 	$enable_debuginfod_question "y"
  

Patch

diff --git a/gdb/linespec.c b/gdb/linespec.c
index e9339c3338c..6db0f02e318 100644
--- a/gdb/linespec.c
+++ b/gdb/linespec.c
@@ -2283,13 +2283,13 @@  convert_linespec_to_sals (struct linespec_state *state, linespec *ls)
 	/* Make sure we have a filename for canonicalization.  */
 	if (ls->explicit_loc.source_filename == NULL)
 	  {
-	    const char *fullname = symtab_to_fullname (state->default_symtab);
+	    const char *filename = state->default_symtab->filename;
 
 	    /* It may be more appropriate to keep DEFAULT_SYMTAB in its symtab
 	       form so that displaying SOURCE_FILENAME can follow the current
 	       FILENAME_DISPLAY_STRING setting.  But as it is used only rarely
 	       it has been kept for code simplicity only in absolute form.  */
-	    ls->explicit_loc.source_filename = xstrdup (fullname);
+	    ls->explicit_loc.source_filename = xstrdup (filename);
 	  }
     }
   else
diff --git a/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp b/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp
index 6d5af1370b0..bfe6e639e4f 100644
--- a/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp
+++ b/gdb/testsuite/gdb.debuginfod/fetch_src_and_symbols.exp
@@ -224,6 +224,15 @@  proc_with_prefix local_url { } {
     set lineno [gdb_get_line_number "Breakpoint here"]
     gdb_test "list $lineno" "return 0;\[^\r\n\]+Breakpoint here\\. .*"
 
+    # Verify that a breakpoint re-sets correctly when the actual location
+    # of the source file in the debuginfod client cache differs from
+    # the contents of DW_AT_comp_dir and DW_AT_name.
+    gdb_test "set cwd $debugdir" "" "file [file tail $binfile] cwd"
+    gdb_test_no_output "del breakpoint 1"
+    gdb_test "break $lineno" "Breakpoint 2 at.*file.*"
+    gdb_test "run" "Breakpoint 2.*" \
+	"file [file tail $binfile] set breakpoint"
+
     # GDB should now find the executable file.
     clean_restart
     gdb_test "core $::corefile" ".*return 0.*" "file [file tail $::corefile]" \