|Summary:||ASTERISK-04411: [patch] pciradio.c does not compile|
|Date Opened:||2005-06-14 04:05:53||Date Closed:||2011-06-07 14:02:46|
|Environment:||Attachments:||( 0) pciradio.patch.txt|
|Description:||pciradio.c does not compile as of CVS 2005-06-14, on Fedora Core 4 (kernel-2.6.11-1.1369_FC4, gcc-4.0.0-8).|
See attached patch to fix this. flags should be long, not int. int flags causes save_flags(&flags) to be compiled to pushfq; popq %edx, the latter not being a valid x86_64 instruction.
|Comments:||By: Tilghman Lesher (tilghman) 2005-06-14 08:36:02|
Where did you get your radio board, or did you build it yourself?
By: sailer (sailer) 2005-06-14 08:48:07
I don't have such a card, I just tried to compile zaptel and it bombed
By: Tilghman Lesher (tilghman) 2005-06-14 11:41:50
I'm kind of nervous about any fix for hardware drivers which is not tested on the hardware itself. Possibly a better solution would be to make the compilation of the pciradio driver optional (in the Makefile). That will solve your issue, without changing the currently-working driver.
By: sailer (sailer) 2005-06-14 12:48:50
I thought the idea is to fix bugs when they are found, instead of stacking kludge over workaround? Especially in a development version?
save_flags and friends is clearly documented to take a long flags, see for example linux/Documentation/cli-sti-removal.txt and linux/Documentation/spinlocks.txt.
By: Tilghman Lesher (tilghman) 2005-06-14 13:50:49
If you can test it on the hardware, I'll consider it a fix. If not, then your patch is a kludge that just happens to compile on your platform.
As you said, we want fixes, not kludges.
By: Mark Spencer (markster) 2005-06-16 20:55:42
Fixed in CVS head.
By: Michael Jerris (mikej) 2005-06-17 06:44:38
pciradio.c does not exist in 1-0 branch