[PATCH v3] [PR gdb/17315] fix until behavior with trailing !is_stmt lines

Andrew Burgess aburgess@redhat.com
Fri Feb 11 15:17:11 GMT 2022


* Bruno Larsen via Gdb-patches <gdb-patches@sourceware.org> [2022-02-10 10:46:01 -0300]:

> On 2/10/22 09:44, Andrew Burgess wrote:
> > * Bruno Larsen via Gdb-patches <gdb-patches@sourceware.org> [2022-01-26 10:08:13 -0300]:
> > 
> > > When using the command "until", it is expected that GDB will exit a
> > > loop if the current instruction is the last one related to that loop.
> > > However, if there were trailing non-statement instructions, "until"
> > > would just behave as "next".  This was most noticeable in clang-compiled
> > > code, but might happen with gcc-compiled as well.  PR gdb/17315 relates
> > > to this problem, as running gdb.base/watchpoint.exp with clang
> > > would fail for this reason.
> > > 
> > > The current patch fixes this by adding an extra check to the
> > > until_next_command function, going through all the following
> > > instructions: if the next instruction relates to the same line and is
> > > not marked as a statement, the end of the execution range is moved to
> > > the end of this next instruction.
> > > 
> > > This patch also adds a test case that can be run with gcc to test that
> > > this functionality is not silently broken in future updates.
> > 
> > Hi Bruno!
> > 
> > Thanks for working on this.  I had some comments, see inline below.
> > 
> 
> Hi Andrew,
> 
> Thanks for the review. I agree with your comments, and I think that most of the explanations are indeed better, with one exception mentioned below.
> 
> > > ---
> > >   gdb/infcmd.c                                  |  23 ++++
> > >   gdb/testsuite/gdb.base/until-trailing-insns.c |  53 ++++++++
> > >   .../gdb.base/until-trailing-insns.exp         | 118 ++++++++++++++++++
> > >   3 files changed, 194 insertions(+)
> > >   create mode 100644 gdb/testsuite/gdb.base/until-trailing-insns.c
> > >   create mode 100644 gdb/testsuite/gdb.base/until-trailing-insns.exp
> > > 
> > > diff --git a/gdb/infcmd.c b/gdb/infcmd.c
> > > index 9f4ed8bff13..5e57437a4cb 100644
> > > --- a/gdb/infcmd.c
> > > +++ b/gdb/infcmd.c
> > > @@ -1346,6 +1346,29 @@ until_next_command (int from_tty)
> > >         tp->control.step_range_start = BLOCK_ENTRY_PC (SYMBOL_BLOCK_VALUE (func));
> > >         tp->control.step_range_end = sal.end;
> > > +
> > > +      /* By setting the step_range_end based on the current pc, we implicitly
> > > +	 assume that the last instruction for any given line must have is_stmt set.
> > > +	 This is not necessarily true.  Clang-13, for example, would compile
> > > +	 the following code:
> > > +
> > > +for(int i=0; i<10; i++)
> > > +  {
> > > +    foo();
> > > +  }
> > 
> > Code samples like this should be indented to the same level as the
> > comment, and now kept on the left.  Also, unless there's good reason
> > not too, code samples within comments should follow the same GNU
> > standard as the rest of GDB, so white space before '(' and around '='
> > and '<' for example.
> > 
> > > +
> > > +	 with 2 entries after the last is_stmt linetable entry.  To fix this, we
> > 
> > You have trailing white space here.
> > 
> > > +	 iterate over the instruction related to the end PC, until we find an sal
> > > +	 related to a different line, and set that pc as the step_range_end. */
> > 
> > Two spaces before '*/'.
> > 
> > > +
> > > +      struct symtab_and_line final_sal;
> > > +      final_sal = find_pc_line (tp->control.step_range_end, 0);
> > > +
> > > +      while (final_sal.line == sal.line && !final_sal.is_stmt)
> > 
> > I think we should also be checking final_sal.symtab == sal.symtab
> > here.  If we find a line for the same line number but from a different
> > symtab (unlikely as that may be) we should probably not step over that
> > code.
> > 
> > > +        {
> > > +	  tp->control.step_range_end = final_sal.end;
> > > +	  final_sal = find_pc_line (final_sal.end, 0);
> > > +        }
> > >       }
> > >     tp->control.may_range_step = 1;
> > > diff --git a/gdb/testsuite/gdb.base/until-trailing-insns.c b/gdb/testsuite/gdb.base/until-trailing-insns.c
> > > new file mode 100644
> > > index 00000000000..68a2033f517
> > > --- /dev/null
> > > +++ b/gdb/testsuite/gdb.base/until-trailing-insns.c
> > > @@ -0,0 +1,53 @@
> > > +/* Copyright 2022 Free Software Foundation, Inc.
> > > +
> > > +   This program is free software; you can redistribute it and/or modify
> > > +   it under the terms of the GNU General Public License as published by
> > > +   the Free Software Foundation; either version 3 of the License, or
> > > +   (at your option) any later version.
> > > +
> > > +   This program is distributed in the hope that it will be useful,
> > > +   but WITHOUT ANY WARRANTY; without even the implied warranty of
> > > +   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> > > +   GNU General Public License for more details.
> > > +
> > > +   You should have received a copy of the GNU General Public License
> > > +   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
> > > +
> > > +/* This test aims to recreate clang-13 behavior in a more controllable
> > > +   environment.  We want to create a linetable like so:
> > > +
> > > +   line | code    | is_stmt |
> > > +    1     i = 0;	Y
> > > +    2     while(1){	N
> > > +    3       if(...)	Y
> > > +    4       a = i;	Y
> > > +    5       i++;	Y
> > > +    6     }		N
> > > +
> > > +    where all lines, with the exception of line 4, all point to the loop code
> > > +    on line 2, effectively creating a "for" loop.  GDB's until command had a
> > > +    problem with this setup, because when we're stopped at line 5 the range
> > > +    of execution would not include line 6 - a jump instruction still related
> > > +    to line 2, making us stop at the next instruction marked as a statement
> > > +    and not related to the current line (in this example, line 4).  This test
> > > +    exercises the command "until" in this complicated situation.
> > > +
> > > + */
> > 
> > The .exp file should include a comment explaining what the test does,
> > so I think this comment would be better moved to there.
> > 
> > > +
> > > +int main(){/* main prologue */
> > > +    asm("main_label: .globl main_label"); /* This is required */
> > > +    asm("loop_start: .globl loop_start");
> > > +    int a, i;
> > > +    i = 0;						/* loop assignment */
> > > +    while (1) {						/* loop line */
> > > +	asm("loop_condition: .globl loop_condition");
> > > +	if (i >= 10) break;				/* loop condition */
> > > +	asm("loop_code: .globl loop_code");
> > > +	a = i;						/* loop code */
> > > +	asm("loop_increment: .globl loop_increment");
> > > +	i ++;						/* loop increment */
> > > +	asm("loop_jump: .globl loop_jump");
> > > +    }
> > > +    asm("main_return: .globl main_return");
> > > +    return 0; /* main return */
> > > +}
> > 
> > This code would ideally follow the GNU coding standard as much as possible.
> > 
> > > diff --git a/gdb/testsuite/gdb.base/until-trailing-insns.exp b/gdb/testsuite/gdb.base/until-trailing-insns.exp
> > > new file mode 100644
> > > index 00000000000..ea341ccd5a5
> > > --- /dev/null
> > > +++ b/gdb/testsuite/gdb.base/until-trailing-insns.exp
> > > @@ -0,0 +1,118 @@
> > > +# Copyright 2022 Free Software Foundation, Inc.
> > > +
> > > +# This program is free software; you can redistribute it and/or modify
> > > +# it under the terms of the GNU General Public License as published by
> > > +# the Free Software Foundation; either version 3 of the License, or
> > > +# (at your option) any later version.
> > > +#
> > > +# This program is distributed in the hope that it will be useful,
> > > +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> > > +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> > > +# GNU General Public License for more details.
> > > +#
> > > +# You should have received a copy of the GNU General Public License
> > > +# along with this program.  If not, see <http://www.gnu.org/licenses/>.
> > > +load_lib dwarf.exp
> > > +
> > > +# This test can only be run on targets which support DWARF-2 and use gas.
> > > +if {![dwarf2_support]} {
> > > +    unsupported "dwarf2 support required for this test"
> > > +    return 0
> > > +}
> > > +if [get_compiler_info] {
> > > +    return -1
> > > +}
> > > +# dwarf assembler requires gcc compiled.
> > > +if !$gcc_compiled {
> > > +    unsupported "gcc is required for this test"
> > > +	return 0
> > > +}
> > > +
> > > +standard_testfile .c .S
> > > +
> > > +if { [prepare_for_testing "failed to prepare" ${testfile} ${srcfile}] } {
> > > +    return -1
> > > +}
> > > +
> > > +
> > > +set asm_file [standard_output_file $srcfile2]
> > > +Dwarf::assemble $asm_file {
> > > +    global srcdir subdir srcfile
> > > +    declare_labels integer_label L
> > > +    set int_size [get_sizeof "int" 4]
> > > +
> > > +    # Find start address and length for our functions.
> > > +    lassign [function_range main [list ${srcdir}/${subdir}/$srcfile]] \
> > > +	main_start main_len
> > > +    set main_end "$main_start + $main_len"
> > > +
> > > +    cu {} {
> > > +	compile_unit {
> > > +	    {language @DW_LANG_C}
> > > +	    {name until-trailing-isns.c}
> > > +	    {stmt_list $L DW_FORM_sec_offset}
> > > +	    {low_pc 0 addr}
> > > +	} {
> > > +	    subprogram {
> > > +		{external 1 flag}
> > > +		{name main}
> > > +		{low_pc $main_start addr}
> > > +		{high_pc $main_len DW_FORM_data4}
> > > +	    }
> > > +	}
> > > +    }
> > > +
> > > +    lines {version 2 default_is_stmt 1} L {
> > > +	include_dir "${srcdir}/${subdir}"
> > > +	file_name "$srcfile" 1
> > > +
> > > +	# Generate a line table program.  This mimicks clang-13's behavior
> > > +	# of adding some !is_stmt at the end of a loop line, making until
> > > +	# not work properly.
> > > +	program {
> > > +	    {DW_LNE_set_address $main_start}
> > > +	    {line [gdb_get_line_number "main prologue"]}
> > > +	    {DW_LNS_copy}
> > > +	    {DW_LNE_set_address loop_start}
> > > +	    {line [gdb_get_line_number "loop line"]}
> > > +	    {DW_LNS_copy}
> > > +	    {DW_LNE_set_address loop_condition}
> > > +	    {line [gdb_get_line_number "loop line"]}
> > > +	    {DW_LNS_negate_stmt}
> > > +	    {DW_LNS_copy}
> > > +	    {DW_LNE_set_address loop_code}
> > > +	    {line [gdb_get_line_number "loop code"]}
> > 
> > I found that this test was failing for me.  The reason being that
> > "loop code" actually appears inside the comment within the .c file.  I
> > found that if I delete the comment in the .c file then the test
> > started passing again (only with your fix of course).
> > 
> > > +	    {DW_LNS_negate_stmt}
> > > +	    {DW_LNS_copy}
> > > +	    {DW_LNE_set_address loop_increment}
> > > +	    {line [gdb_get_line_number "loop line"]}
> > > +	    {DW_LNS_copy}
> > > +	    {DW_LNE_set_address loop_jump}
> > > +	    {line [gdb_get_line_number "loop line"]}
> > > +	    {DW_LNS_negate_stmt}
> > > +	    {DW_LNS_copy}
> > > +	    {DW_LNE_set_address main_return}
> > > +	    {line [gdb_get_line_number "main return"]}
> > > +	    {DW_LNS_negate_stmt}
> > > +	    {DW_LNS_copy}
> > > +	    {DW_LNE_set_address $main_end}
> > > +	    {line [expr [gdb_get_line_number "main return"] + 1]}
> > > +	    {DW_LNS_copy}
> > > +	    {DW_LNE_end_sequence}
> > > +	}
> > > +    }
> > > +
> > > +}
> > > +
> > > +if { [prepare_for_testing "failed to prepare" ${testfile} \
> > > +	[list $srcfile $asm_file] {nodebug} ] } {
> > > +    return -1
> > > +}
> > > +
> > > +if ![runto_main] {
> > > +    return -1
> > > +}
> > > +
> > > +gdb_test "next" ".* loop code .*" "inside the loop"
> > > +gdb_test "next" ".* loop line .*" "ending of loop"
> > > +gdb_test "until" ".* return 0; .*" "left loop"
> > 
> > While reviewing this patch I ended up making the changes I recommended
> > above as I went along.
> > 
> > I also rewrote and extended some of the text describing the problem we
> > are hitting.  I always find these things easier to understand if there
> > is a really simple worked example included, so that's what I did.
> > 
> > I've also added a 'Bug: ' link within the commit message.
> > 
> > Could you take a look at the patch below, if you are happy with it
> > then I propose that we merge this.
> > 
> > Thanks,
> > Andrew
> > 
> > ---
> > 
> > commit 209580c6befd577b2946a09bc2e024d10d7f0a54
> > Author: Bruno Larsen via Gdb-patches <gdb-patches@sourceware.org>
> > Date:   Wed Jan 26 10:08:13 2022 -0300
> > 
> >      gdb: fix until behavior with trailing !is_stmt lines
> >      When using the command "until", it is expected that GDB will exit a
> >      loop if the current instruction is the last one related to that loop.
> >      However, if there were trailing non-statement instructions, "until"
> >      would just behave as "next".  This was noticeable in clang-compiled
> >      code, but might happen with gcc-compiled as well.  PR gdb/17315 relates
> >      to this problem, as running gdb.base/watchpoint.exp with clang
> >      would fail for this reason.
> >      To better understand this issue, consider the following source code,
> >      with line numbers marked on the left:
> >        10:   for (i = 0; i < 10; ++i)
> >        11:     loop_body ();
> >        12:   other_stuff ();
> >      If we transform this to pseudo-assembler, and generate a line table,
> >      we could end up with something like this:
> >        Address | Pseudo-Assembler | Line | Is-Statement?
> >        0x100   | i = 0            | 10   | Yes
> >        0x104   | loop_body ()     | 11   | Yes
> >        0x108   | i = i + 1        | 10   | Yes
> >        0x10c   | if (i < 10):     | 10   | No
> >        0x110   |     goto 0x104   | 10   | No
> >        0x114   | other_stuff ()   | 12   | Yes
> >      Notice the two non-statement instructions at the end of the loop.
> >      The problem is that when we reach address 0x108 and use 'until',
> >      hoping to leave the loop, GDB sets up a stepping range that runs from
> >      the start of the function (0x100 in our example) to the end of the
> >      current line table entry, that is 0x10c in our example.  GDB then
> >      starts stepping forward.
> >      When 0x10c is reached GDB spots that we have left the stepping range,
> >      that the new location is not a statement, and that the new location is
> >      associated with the same source line number as the previous stepping
> >      range.  GDB then sets up a new stepping range that runs from 0x10c to
> >      0x114, and continues stepping forward.
> >      Within that stepping range the inferior hits the goto (at 0x110) and
> >      loops back to address 0x104.
> >      At 0x104 GDB spots that we have left the previous stepping range, that
> >      the new address is marked as a statement, and that the new address is
> >      for a different source line.  As a result, GDB stops and returns
> >      control to the user.  This is not what the user was expecting, they
> >      expected GDB to exit the loop.
> >      The fix proposed in this patch, is that, when the user issues the
> >      'until' command, and GDB sets up the initial stepping range, GDB will
> >      check subsequent SALs (symtab_and_lines) to see if they are
> >      non-statements associated with the same line number.  If they are then
> >      the end of the initial stepping range is extended to the end of the
> >      non-statement SALs.
> >      In our example above, the user is at 0x108 and uses 'until', GDB now
> >      sets up a stepping range from the start of the function 0x100 to
> >      0x114, the first address associated with a different line.
> >      Now as GDB steps around the loop it never leaves the initial stepping
> >      range.  It is only when GDB exits the loop that we leave the stepping
> >      range, and the stepping finishes at address 0x114.
> >      This patch also adds a test case that can be run with gcc to test that
> >      this functionality is not broken in the future.
> >      Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=17315
> > 
> > diff --git a/gdb/infcmd.c b/gdb/infcmd.c
> > index 48dda9b23ee..a25d024d24c 100644
> > --- a/gdb/infcmd.c
> > +++ b/gdb/infcmd.c
> > @@ -1346,6 +1346,45 @@ until_next_command (int from_tty)
> >         tp->control.step_range_start = BLOCK_ENTRY_PC (SYMBOL_BLOCK_VALUE (func));
> >         tp->control.step_range_end = sal.end;
> > +
> > +      /* By setting the step_range_end based on the current pc, we are
> > +	 assuming that the last line table entry for any given source line
> > +	 will have is_stmt set to true.  This is not necessarily the case,
> > +	 there may be additional entries for the same source line with
> > +	 is_stmt set false.  Consider the following code:
> > +
> > +	 for (int i = 0; i < 10; i++)
> > +	   loop_body ();
> > +
> > +	 Clang-13, will generate multiple line table entries at the end of
> > +	 the loop all associated with the 'for' line.  The first of these
> > +	 entries is marked is_stmt true, but the other entries are is_stmt
> > +	 false.
> > +
> > +	 If we only use the values in SAL then our stepping range will not
> > +	 extend beyond the loop and the until command will exit the
> > +	 stepping range, and find the is_stmt true location corresponding
> > +	 to 'loop_body ()', this will cause GDB to stop inside the loop on
> > +	 the which is not what we wanted.
> 
> The wording in this paragraph seems a bit odd to me. My suggestion for an improvement is:
> 
> If we only use the values in SAL, then our stepping range may not
> extend to the end of the loop. The until command will reach the end
> of the range, find a non is_stmt instruction, and step to the next
> is_stmt instruction. This stopping point, however, will be inside
> the loop, which is not what we wanted.

I made this change, and pushed the patch.

Thanks,
Andrew



More information about the Gdb-patches mailing list