On 09/06/2017 06:51 AM, Juan Quintela wrote: > So far, we had to free the error after each caller, so just do it > here. Once there, tls.c was leaking the error. You mention tls.c, > > Signed-off-by: Juan Quintela > --- > migration/channel.c | 1 - > migration/migration.c | 10 ++++------ > migration/migration.h | 4 ++-- > migration/socket.c | 1 - > 4 files changed, 6 insertions(+), 10 deletions(-) but don't touch it. Am I missing something? > > -void migrate_fd_error(MigrationState *s, const Error *error) > +void migrate_fd_error(MigrationState *s, Error *error) > { No comments at definition, > +++ b/migration/migration.h > @@ -163,8 +163,8 @@ bool migration_has_all_channels(void); > > uint64_t migrate_max_downtime(void); > > -void migrate_set_error(MigrationState *s, const Error *error); > -void migrate_fd_error(MigrationState *s, const Error *error); > +void migrate_set_error(MigrationState *s, Error *error); > +void migrate_fd_error(MigrationState *s, Error *error); or at declaration. That would be worth adding at some point, but this patch isn't making it worse. The code looks okay in isolation, so if it is only the commit message that needs fixing, Reviewed-by: Eric Blake -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org