On 2007-08-10 at 19:13:26 [+0200], Ingo Weinhold <bonefish@xxxxxxxxxxxxxxx> wrote: > On 2007-08-09 at 17:31:20 [+0200], Jonas Sundström <jonas@xxxxxxxxxxx> > wrote: > > I just saw a Haiku sh -login process without a rld_lock > > semaphore, only having the fork_lock and the env_lock. > > > > Is that supposed to happen? > > Good catch. I guess, the runtime loader needs to register an fork hook, or > the child fork initialization. I shouldn't write mails, when I'm too tired... Anyway, the runtime loader is explicitely asked to reinitialize itself after fork(), now. > It also doesn't look quit right, that the __init_fork(), which initializes > a static sForkLock variable -- is called by fork(), but the variable is > already used by atfork(). The code is OK, __init_fork() is already called in the libroot initialization. CU, Ingo