Re: [PATCH 1/2] Fix reverse stepping multiple contiguous PC ranges over the line table.

[Simon Marchi <simark@simark.ca>]

On 6/22/23 12:52, Carl Love wrote:
> 
> Simon:
> 
> On Mon, 2023-06-19 at 13:11 -0400, Simon Marchi wrote:
>>> --- a/gdb/testsuite/lib/gdb.exp
>>> +++ b/gdb/testsuite/lib/gdb.exp
>>> @@ -4794,6 +4794,8 @@ proc quote_for_host { args } {
>>>   #     debug information
>>>   #   - text_segment=addr: Tell the linker to place the text
>>> segment at ADDR.
>>>   #   - build-id: Ensure the final binary includes a build-id.
>>> +#   - no-column-info: Disable generation of column table
>>> information.
>>> +#   - column-info: Enable generation of column table information.
>>>   #
>>>   # And here are some of the not too obscure options understood by
>>> DejaGnu that
>>>   # influence the compilation:
>>> @@ -5003,6 +5005,34 @@ proc gdb_compile {source dest type options}
>>> {
>>>               } else {
>>>                   error "Don't know how to handle text_segment
>>> option."
>>>               }
>>> +     } elseif { $opt == "column-info" } {
>>> +         if {[test_compiler_info {gcc-*}]} {
>>> +             if {[test_compiler_info {gcc-[1-6]-*}]} {
>>> +                 error "gdb_compile option no-column-info not
>>> supported."
>>
>> I think this path should return the equivalent of "failed to
>> compile",
>> instead of throwing an error.  Control will go back to the test,
>> which
>> will generally skip the portion of the test that requires that
>> binary.
> 
> Not entirely sure how to accomplish what you are looking for.
> 
> I change:
>   error "gdb_compile option no-column-info not supported."
> to
>   set result "option no-column-info not supported."
>   clone_output "gdb compile failed, $result"
>   return 1
> 
> When I force the if {[test_compiler_info...]} tp be true to test this,
> I get:
> 
>    get_compiler_info: gcc-12-2-1
>    gdb compile failed, option no-column-info not supported.
>    UNTESTED: gdb.reverse/func-map-to-same-line.exp:
>    with_column_info=yes: failed t\
>    o prepare
>    testcase /home/carll/GDB/build-reverse-multiple-
>    contiguous/gdb/testsuite/../../\
>    ../binutils-gdb-reverse-multiple-
>    contiguous/gdb/testsuite/gdb.reverse/func-map-\
>    to-same-line.exp completed in 0 seconds
> 
>                    === gdb Summary ===
> 
>    # of untested testcases         1
> 
> The test case doesn't have any part of the test that doesn't require
> compiling so it is not clear if that would work with this fix.  Anyway,
> wanted to run that by you to see if this is an appropriate fix?  I am
> really not sure about it.  Thanks.

I think that's the expected behavior.  The UNTESTED is emitted by
build_executable_from_specs, I think.  If the test used gdb_compile, I
think we wouldn't see an UNTESTED.  But as far as your addition is
concerned, I think it's fine.

I just thought of a simpler alternative though.  Just remove the version
check.  If we build with an older gcc, there will simply be a message
that says that the flag is not recognized, and the result should be just
the same.  I just hacked it locally and changed the flag name to be
wrong (I don't have a gcc <= 6 on hand to test).  It looks like:

    Executing on host: gcc  -fno-stack-protector  -fdiagnostics-color=never -gcolumn-info-foo -c -g  -o /home/simark/build/binutils-gdb/gdb/testsuite/outputs/gdb.reverse/func-map-to-same-line/func-map-to-same-line0.o /home/simark/src/binutils-gdb/gdb/testsuite/gdb.reverse/func-map-to-same-line.c    (timeout = 300)
    builtin_spawn -ignore SIGHUP gcc -fno-stack-protector -fdiagnostics-color=never -gcolumn-info-foo -c -g -o /home/simark/build/binutils-gdb/gdb/testsuite/outputs/gdb.reverse/func-map-to-same-line/func-map-to-same-line0.o /home/simark/src/binutils-gdb/gdb/testsuite/gdb.reverse/func-map-to-same-line.c

    gcc: error: unrecognized debug output level 'column-info-foo'

    compiler exited with status 1
    output is:
    gcc: error: unrecognized debug output level 'column-info-foo'


    gdb compile failed, gcc: error: unrecognized debug output level 'column-info-foo'
    UNTESTED: gdb.reverse/func-map-to-same-line.exp: with_column_info=yes: failed to prepare

I then thought about the "no-column-info" case.  Currently, you error
out for gccs <= 6.  However, shouldn't we just compile without any
special flag in that case, since there just wasn't any support for
column-info back then?

Simon