Guys -
I get similar problems building USRP (not yet a Debian package).
I can fix things up by deleting the line
#include linux/compat.h
from /usr/include/linux/usbdevice_fs.h.
This change is not obviously broken, as it is one of only two *.h
files in there that includes compat.h. The other one is kexec.h,
which looks to me like a unique special case.
After making that change, I can confirm that fxload and openct
build on amd64 again.
I have no idea what form is desired for patch submissions, but
I'll try with an attachment. This patch makes the deletion I
described above, and adds a regression test for usbdevice_fs.h
to the testsuite directory.
I hope this helps.
- Larry
diff -urN ../linux-kernel-headers-2.6.13+0rc3/include/linux/usbdevice_fs.h
./include/linux/usbdevice_fs.h
--- ../linux-kernel-headers-2.6.13+0rc3/include/linux/usbdevice_fs.h
2005-07-12 21:46:46.0 -0700
+++ ./include/linux/usbdevice_fs.h 2005-09-13 10:36:55.0 -0700
@@ -32,7 +32,6 @@
#define _LINUX_USBDEVICE_FS_H
#include linux/types.h
-#include linux/compat.h
/* - */
diff -urN ../linux-kernel-headers-2.6.13+0rc3/testsuite/Makefile
./testsuite/Makefile
--- ../linux-kernel-headers-2.6.13+0rc3/testsuite/Makefile 2005-09-13
10:48:26.0 -0700
+++ ./testsuite/Makefile2005-09-13 10:33:16.0 -0700
@@ -2,7 +2,7 @@
TESTS_295 = $(patsubst %,295-%,$(TESTS))
# Filter out tests which no one tries to use -ansi; not worth fixing.
-NON_ANSI = videodev.o videodev-time.o cdrom.o
+NON_ANSI = videodev.o videodev-time.o cdrom.o usbdevice_fs.o
TESTS_ANSI = $(patsubst %,ansi-%,$(filter-out $(NON_ANSI),$(TESTS)))
# Similarly for C++.
diff -urN ../linux-kernel-headers-2.6.13+0rc3/testsuite/usbdevice_fs.c
./testsuite/usbdevice_fs.c
--- ../linux-kernel-headers-2.6.13+0rc3/testsuite/usbdevice_fs.c
1969-12-31 16:00:00.0 -0800
+++ ./testsuite/usbdevice_fs.c 2005-09-13 10:25:19.0 -0700
@@ -0,0 +1,6 @@
+#include linux/usbdevice_fs.h
+
+int main()
+{
+ return 0;
+}
signature.asc
Description: Digital signature