Hi
I found a strange behaviour in `dlerror` with regard
to errno. When the function is called, it returns the
correct error message but sets errno to ENOSYS.
Thus, future usages of errno don't reflect the real reason
why the previous call to `dlopen` failed.
A little test case for this quirk is
* libddekit/interrupt.c (irq_server): If a message with a valid reply
port arrives, deallocate that instead of calling 'device_intr_enable'
to re-enable the interrupt.
---
libddekit/interrupt.c | 6 +-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/libddekit/interrupt.c b/libddek
Justus Winter, on Sun 28 Feb 2016 00:33:21 +0100, wrote:
> * i386/Makefrag.am (libkernel_a_SOURCES): Add new file.
> * i386/i386/cpu.h: New file.
Ack, thanks.
> ---
> i386/Makefrag.am | 1 +
> i386/i386/cpu.h | 110
> +++
> 2 files changed,
* i386/Makefrag.am (libkernel_a_SOURCES): Add new file.
* i386/i386/cpu.h: New file.
---
i386/Makefrag.am | 1 +
i386/i386/cpu.h | 110 +++
2 files changed, 111 insertions(+)
create mode 100644 i386/i386/cpu.h
diff --git a/i386/Makefrag.am b
Hello,
I revised the patch that is delegating interrupts to userspace.
Notable changes:
1/ We now use proper dead-name notifications to detect that the send
right we use to send notifications goes stale, i.e. the userspace
server dies.
2/ 'device_intr_enable' is a nop now...
3/ ... instead the
---
Makefrag.am | 9 ++
device/ds_routines.c | 27
device/interrupt.h | 34
device/intr.c| 341 +++
include/device/intr.h| 18 +++
include/mach/experimental.defs | 9
