Re: [PATCH 3/3] Add -file-list-shared-libraries MI command

["Marc-André Laperle" <marc-andre.laperle@ericsson.com>]

Thanks for the comments! A few questions/suggestions below.


On 2016-11-23 08:06 AM, Pedro Alves wrote:
> Hi Marc-Andre,
>
> Great that you're tackling this.  Thanks much for digging into GDB.  :-)
> I like this.  As for following GDB's code standards, it's almost perfect.
>
> But I have questions on the MI output, though.  See below.
>
> On 09/12/2016 09:27 PM, Marc-Andre Laperle wrote:
>
>> diff --git a/gdb/doc/gdb.texinfo b/gdb/doc/gdb.texinfo
>> index d1a5e7c..c6b2133 100644
>> --- a/gdb/doc/gdb.texinfo
>> +++ b/gdb/doc/gdb.texinfo
>> @@ -31401,26 +31401,35 @@ The @value{GDBN} equivalent is @samp{info sources}.
>>   (gdb)
>>   @end smallexample
>>   
>> -@ignore
>>   @subheading The @code{-file-list-shared-libraries} Command
>>   @findex -file-list-shared-libraries
>>   
>>   @subsubheading Synopsis
>>   
>>   @smallexample
>> - -file-list-shared-libraries
>> + -file-list-shared-libraries [ @var{regexp} ]
>>   @end smallexample
>>   
>>   List the shared libraries in the program.
>> +With a regular expression @var{regexp}, only those libraries whose
>> +names match @var{regexp} are listed.
>>   
>>   @subsubheading @value{GDBN} Command
>>   
>>   The corresponding @value{GDBN} command is @samp{info shared}.
>>   
>>   @subsubheading Example
>> -N.A.
>> +@smallexample
>> +(gdb)
>> +-file-list-exec-source-files
>> +^done,shared-libraries=[
>> +@{from="0x72815989",to="0x728162c0",syms-read="1",name="/lib/libfoo.so"@},
>> +@{from="0x76ee48c0",to="0x76ee9160",syms-read="1",name="/lib/libbar.so"@}]
> I don't see where the documentation describes what the
> attributes of each list element are.  Sorry if I missed it.
>
> I find it surprising that the attributes output don't match
> attributes output by =library-loaded ?  I'd think they should
> match.  That'd simplify the documentation too, as one place
> would refer to the other for attributes list.
>
> BTW, =library-loaded doesn't output from/to.  ISTR that that
> was discussed and left out, because from/to assume there's
> a contiguous range to report, while that's not true on all
> targets (i.e., assumes a single segment).  So it seems to
> me that that's a separate discussion/patch would better
> be addressed in both places.

Those attributes are from the "info shared" command which is what this 
new MI command is inspired from. The from/to are shown in Eclipse and I 
personally use it to know in which library a given stack frame is, in 
absence of debug symbols. I could change the attributes to be just like 
"=library-loaded" but it would be inconsistent with "info shared" and 
will remove information that was available before. Perhaps the best way 
forward would be to make all attributes the same as "=library-loaded" 
but also add the from/to.
To make the MI more future proof, it could be a list of from/to segments 
instead, but for now it will only one segment. What do you think?

>> +void
>> +mi_cmd_file_list_shared_libraries (char *command, char **argv, int argc)
>> +{
>> +  struct ui_out *uiout = current_uiout;
>> +  const char *pattern;
>> +  struct so_list *so = NULL;
>> +  struct gdbarch *gdbarch = target_gdbarch ();
>> +
>> +  switch (argc)
>> +    {
>> +    case 0:
>> +      pattern = NULL;
>> +      break;
>> +    case 1:
>> +      pattern = argv[0];
>> +      break;
>> +    default:
>> +      error (_("Usage: -file-list-shared-libraries [REGEXP]"));
>> +      break;
> error doesn't return.  No need for the break.
>
>> +    }
>> +
>> +  if (pattern != NULL)
>> +    {
>> +      char *re_err = re_comp (pattern);
>> +
>> +      if (re_err != NULL)
>> +	error (_("Invalid regexp: %s"), re_err);
>> +    }
>> +
>> +  update_solib_list (1);
>> +
>> +  /* Print the table header.  */
>> +  ui_out_begin (uiout, ui_out_type_list, "shared-libraries");
>> +
>> +  ALL_SO_LIBS (so)
>> +    {
>> +      if (so->so_name[0] == '\0')
>> +	continue;
>> +      if (pattern != NULL && !re_exec (so->so_name))
>> +	continue;
>> +
>> +      ui_out_begin (uiout, ui_out_type_tuple, NULL);
>> +
>> +      if (so->addr_high != 0)
>> +	{
>> +	  ui_out_field_core_addr (uiout, "from", gdbarch, so->addr_low);
>> +	  ui_out_field_core_addr (uiout, "to", gdbarch, so->addr_high);
>> +	}
>> +      else
>> +	{
>> +	  ui_out_field_skip (uiout, "from");
>> +	  ui_out_field_skip (uiout, "to");
>> +	}
>> +
>> +      ui_out_field_int (uiout, "syms-read", so->symbols_loaded ? 1 : 0);
>> +
>> +      ui_out_field_string (uiout, "name", so->so_name);
> So seems to me that the inner body of this loop would be better
> calling a function that is shared with =library-loaded.
>
>> +
>> --- a/gdb/solist.h
>> +++ b/gdb/solist.h
>> @@ -23,6 +23,11 @@
>>   /* For domain_enum domain.  */
>>   #include "symtab.h"
>>   
>> +#define ALL_SO_LIBS(so) \
>> +    for (so = current_program_space->so_list; \
>> +	 so; \
> Write explicit 'so != NULL'.
>
>> +	 so = so->next)
>> +
>>   /* Forward declaration for target specific link map information.  This
>>      struct is opaque to all but the target specific file.  */
>>   struct lm_info;
>> diff --git a/gdb/testsuite/gdb.mi/mi-solib.exp b/gdb/testsuite/gdb.mi/mi-solib.exp
>> index 2227987..4c40ba4 100644
>> --- a/gdb/testsuite/gdb.mi/mi-solib.exp
>> +++ b/gdb/testsuite/gdb.mi/mi-solib.exp
>> @@ -48,27 +48,47 @@ if { [gdb_compile_shlib ${srcfile_lib} ${binfile_lib} $lib_flags] != ""
>>   
>>   mi_delete_breakpoints
>>   mi_gdb_reinitialize_dir $srcdir/$subdir
>> -mi_gdb_reinitialize_dir $srcdir/$subdir
>>   mi_gdb_load ${binfile}
>>   
>>   mi_load_shlibs $binfile_lib
>>   
>> -mi_gdb_test "777-gdb-set stop-on-solib-events 1" "777\\^done" \
>> -    "set stop-on-solib-events"
>> +proc test_stop_on_solib_events {} {
> Looks like a good candidate for the the new proc_with_prefix.
>
>> +    mi_gdb_test "777-gdb-set stop-on-solib-events 1" "777\\^done" \
>> +	"set stop-on-solib-events"
>>   
>> -# We use "run" rather than "-exec-run" here in order to test that CLI
>> -# commands still cause the correct MI output to be generated.
>> -mi_run_with_cli
>> +    # We use "run" rather than "-exec-run" here in order to test that CLI
>> +    # commands still cause the correct MI output to be generated.
>> +    mi_run_with_cli
>>   
>> -# Also test that the CLI solib event note is output.
>> -set test "CLI prints solib event"
>> -gdb_expect {
>> -    -re "~\"Stopped due to shared library event \\(no libraries added or removed\\)\\\\n" {
>> -	pass "$test"
>> -    }
>> -    timeout {
>> -	fail "$test (timeout)"
>> +    # Also test that the CLI solib event note is output.
>> +    set test "CLI prints solib event"
>> +    gdb_expect {
>> +	-re "~\"Stopped due to shared library event \\(no libraries added or removed\\)\\\\n" {
>> +	    pass "$test"
>> +	}
>> +	timeout {
>> +	    fail "$test (timeout)"
>> +	}
>>       }
>> +
>> +    mi_expect_stop solib-event .* .* .* .* .* "check for solib event"
>> +
>> +    # Unset solib events to avoid interfering with other tests.
>> +    mi_gdb_test "778-gdb-set stop-on-solib-events 0" "778\\^done" \
>> +	"unset stop-on-solib-events"
>> +}
>> +
>> +proc test_file_list_shared_libraries {} {
>> +    global libname
>> +    global binfile
>> +
>> +    mi_continue_to main
>> +
>> +    mi_gdb_test "222-file-list-shared-libraries" \
>> +	"222\\^done,shared-libraries=\\\[.*\{from=\".*\",to=\".*\",syms-read=\"1\",name=\".*${libname}.so\"\}.*]" \
>> +	"Getting a list of shared libraries."
> Lowercase, no period at end, and use imperative:
>
> 	"get list of shared libraries"
>
> Thanks,
> Pedro Alves
>