Re: [PATCH][gdb/testsuite] Replace prepare_for_testing with build_executable

[Tom de Vries via Gdb-patches <gdb-patches@sourceware.org>]

On 9/23/21 5:12 PM, Simon Marchi wrote:
> On 2021-09-17 6:49 p.m., Tom de Vries via Gdb-patches wrote:
>> Hi,
>>
>> I noticed a pattern:
>> ...
>> if {[prepare_for_testing "failed to prepare" $testfile $srcfile $flags]} {
>>     return -1
>> }
>>
>> foreach var {a b} {
>>     clean_restart ${binfile}
>>     ...
>> }
>> ...
>>
>> In this case, gdb is restarted three times, while it's only necessary two
>> times.
>>
>> Fix this by replacing prepare_for_testing with build_executable.
>>
>> I wrote a script that attempts this tranformation on each test-case (only the
>> per-test part shown here):
>> ...
>>         cd $build
>>         make check RUNTESTFLAGS=$test
>>         sum=$build/testsuite/gdb.sum
>>         if ! grep -q "expected passes" $sum; then
>>             echo skipping $test
>>             return
>>         fi
>>
>>         grep ^# $sum > ref.txt
>>         sed -i 's/prepare_for_testing/build_executable/' $src/$test
>>         make check RUNTESTFLAGS=$test
>>
>>         grep ^# $sum > mod.txt
>>         if cmp ref.txt mod.txt; then
>>             return
>>         else
>>             ( cd $src; git ch -f $test )
>>         fi
>> ...
>>
>> This patch is the result of running the script on all test-cases, minus the
>> cases where prepare_for_testing was used in a comment.  Touches 76 test-cases.
> 
> I didn't check all changes, but I agree with the idea of the patch.

Good to hear.

I've reviewed all changes once more.

I found a problem in cpp-linkage-name.exp.  The test passes, but the gdb
commands execute on a lingering gdb instance started by function_range,
so I had to fix that first (by adding gdb_exit at the end of
function_range) to trigger the FAIL introduced by the script.  I've
removed the change in that test-case, and similar cases, and am
currently retesting.

Thanks,
- Tom