|
From: Nicholas N. <n.n...@gm...> - 2009-07-22 08:10:22
|
On Wed, Jul 22, 2009 at 5:32 PM, Bart Van Assche<bar...@gm...> wrote: > > There are two reasons why I added --auto-run-dsymutil=yes: > * At least on the system I ran my tests on, the .dSYM directory was > not built by default. > * The message that suggests to add --auto-run-dsymutil=yes is hard to > filter away from the regression test output because it consists of two > lines of output. I'd prefer to work out why the dSYMs aren't being built for your system. It's an old one -- 9.2.0 or something like that? >> Also, how is DRD's --var-info=yes different to the core's >> --read-var-info=yes? It seems like DRD should just use >> --read-var-info=yes the way Memcheck and Helgrind do. > > If I remember correctly, --var-info was added to DRD before > --read-var-info was added to Memcheck or Helgrind. I have renamed the > DRD option --var-info to --read-var-info and left --var-info as an > undocumented option in DRD. Ah. But after your change there is now two --read-var-info options, one for the core and one for DRD. Which one takes precedence? Maybe both of the booleans get set... but it's not a good situation. The usage message becomes confusing as well. DRD may have had that flag first, but now that the core has it also it would be good if DRD instead used the core's version (VG_(clo_read_var_info)) and didn't look for --read-var-info itself. Nick |