Paolo Mantegazza wrote:
Jan Kiszka wrote:
Hi,
for unknown reasons, someone ;) re-inserted the
EXPORT_SYMBOL(rt_printk) line into hal.c. Unfortunately, as rt_printk
is defined to be printk with adeos, this leads to an export of printk!
Now things get funny: insmod works because all printk symbols in
rtai_hal gets replaced with the kernel ones and the export of rtai_hal
seems to be ignored. But depmod fails as it detects a look in
rtai_hal: that module requires printk while it also exports the
symbol. Nice bug, took some hours to track down...
Attached is a patch for vulcano which removes the bogus exports (only
for x86, don't know what's the case for other archs).
Thanks.
I appled your fixes to Vulcano CVS, but just for hal.c and hal.piped.
Ah, I see. That also explains why you re-added the export. I should have
looked at *both* files thoroughly.
BTW, is there no other way of maintaining such similar files in the
source tree? If there had been some "#ifdef IMMED" or so around the
rt_printk implementation, the wrong export might have been noticed earlier.
Jan
