|
From: Ivo R. <ivo...@gm...> - 2015-08-17 08:12:09
|
Please could you review proposed patch for bug: 351386 Cannot run ld.so.1 under Valgrind (https://bugs.kde.org/show_bug.cgi?id=351386) and my findings posted there? Although the changes are Solaris specific, they touch area shared with Linux so I'd rather double check. Thanks, I. |
|
From: Julian S. <js...@ac...> - 2015-08-17 14:37:03
|
On 17/08/15 10:12, Ivo Raisr wrote: > Please could you review proposed patch for bug: > 351386 Cannot run ld.so.1 under Valgrind > (https://bugs.kde.org/show_bug.cgi?id=351386) > and my findings posted there? Looks fine to me. My only thought is, is info->real_phdr_present always guaranteed to be initialised (to something)? It may be that the whole struct is zeroed out right at the start. Is that so? J |
|
From: Ivo R. <ivo...@gm...> - 2015-08-17 17:05:56
|
2015-08-17 16:36 GMT+02:00 Julian Seward <js...@ac...>: > On 17/08/15 10:12, Ivo Raisr wrote: > > Please could you review proposed patch for bug: > > 351386 Cannot run ld.so.1 under Valgrind > > (https://bugs.kde.org/show_bug.cgi?id=351386) > > and my findings posted there? > > Looks fine to me. My only thought is, is info->real_phdr_present > always guaranteed to be initialised (to something)? It may be that > the whole struct is zeroed out right at the start. Is that so? > Yes, it is. On all OSes it is initialized with VG_(memset)() in load_client() located in corresponding m_initimg/initimg-<os>.c right before VG_(do_exec)() is called. Thank you for review! I. |