Re: pulseaudio mutex bug reported
On 25/11/11 19:02, Bob Tracy wrote:
> On Fri, Nov 25, 2011 at 06:43:06PM +1300, Michael Cree wrote:
>> On 25/11/11 17:25, Bob Tracy wrote:
>>> BEWARE the recent "udev" upgrade: it suffers from the same problem as
>>> the previous version -- the stubbed-out syscalls.
>> The problem is not in udev.
> That's a technicality :-).
> "udev" was built with the include files for
> a version of libc that has the "missing" syscalls stubbed-out. Yes, the
> syscalls are in current kernel.org kernels, but as you point out, until
> we get them added to libc, udev upgrades will continue to be an issue.
> When they get added to libc, I suspect we'll need a new build of udev
> against the new library.
On that point I believe you are mistaken.
It is a little ironic that the maintainer of the Alpha port of glibc
thought that the accept4 syscall was implemented in the kernel so all
the support is in eglibc including the export of accept() for user
programs to link to. The only thing that falls over is that glibc at
some point either realises it does not have the kernel syscall number,
or maybe it calls some unimplemented syscall, but whatever the call
returns back to the calling program with a not-implemented error code.
Once libc6.1 is recompiled with the updated kernel headers it should
start executing the kernel accept4 syscall correctly.
This is an instance where shareable libraries is an advantage. A bug
fixed in the library fixes the symptoms of the bug in all affected packages.