Behavior of the compiler does not match the documentation. The following "-femit-struct-debug-detailed=ind:ord:base" means that debug info should be emitted for structs used indirectly and declared in file that matches the base name of the compilation file (test.c/test.h). So, in the particular case, when struct is used indirectly in test.c and declared in test-nonbase.h - info on structs shouldn't be emitted. In case of, "-femit-struct-debug-detailed=ord:base" - it works correctly and emits nothing. I think that there is a bug in interpretation of optional parameters. /* With optimism, Evgeny Gavrin email : evgeny.gavrin@xxxxxxxxxxx */ ---------------------------------------- > Date: Wed, 19 Jun 2013 11:08:05 -0700 > Subject: Re: -femit-struct-debug-reduced/baseonly debug info > From: iant@xxxxxxxxxx > To: evgeny.gavrin@xxxxxxxxxxx > CC: gcc-help@xxxxxxxxxxx; slava.garbuzov@xxxxxxxxx > > On Wed, Jun 19, 2013 at 9:29 AM, Evgeny Gavrin > <evgeny.gavrin@xxxxxxxxxxx> wrote: >> >> This means that first optional parameter shouldn't affect so much. In case it's omitted debug info should be emitted both for [dir:] and [ind:]. >> >> Two cases: >> 1) test.c -O0 -g2 -femit-struct-debug-detailed=ind:ord:base >> >> opts->x_debug_struct_ordinary[usage] = files; >> >> 2) test.c -O0 -g2 -femit-struct-debug-detailed=ord:base >> >> opts->x_debug_struct_ordinary[DINFO_USAGE_DFN] = files; >> opts->x_debug_struct_ordinary[DINFO_USAGE_DIR_USE] = files; >> opts->x_debug_struct_ordinary[DINFO_USAGE_IND_USE] = files; >> >> What am I missing? On my view, there is smth wrong, or processing of this option has some additional shadow logic, that I failed to find. > > > Are you saying that the behaviour of the compiler does not match the > documentation? Or are you saying that you don't understand why it > works? Or are you saying that the documentation is unclear? > > Ian
