On Thu, Jan 08, 2026 at 03:08:04PM +0530, Prasad Pandit wrote: > On Wed, 7 Jan 2026 at 22:42, Peter Xu <[email protected]> wrote: > >> * Should 'hook_data' pointer be g_autofree too? Where is it free'd > >> otherwise? > > > > hook_data is freed in end_hook(). This patch doesn't change that fact for > > postcopy. It's the smae to non-postcopy tests. > > > >> * The changes look okay; But if tests define hook_data = NULL; Where > >> does it get populated? > > > > It's populated in start_hook() conditionally. When populated, it is always > > (and a must) to be released in end_hook(). > > === > $ grep -Eri 'test_postcopy_common|test_postcopy_recovery_common' > tests/qtest/migration/ -l > tests/qtest/migration/framework.h > tests/qtest/migration/framework.c > tests/qtest/migration/tls-tests.c > tests/qtest/migration/postcopy-tests.c > === > > * Only tls-tests above seem to define and use these hooks properly > along with the hook_data. Postcopy-tests and all other users of > start_hook/end_hook don't seem to use [postcopy|hook]_data at all. Do > we really need this hook_data parameter? Couldn't it be defined as a > tls-tests specific object. (just wondering)
Sorry I don't follow. We need the hook_data for cleaning up tls objects later in end_hook, for either postcopy or other tls tests. > > Reviewed-by: Prasad Pandit <[email protected]> Thanks, > > Thank you. > --- > - Prasad > -- Peter Xu
