From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from mail1.radix.net ([207.192.128.31]:51206 "EHLO mail1.radix.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1761228AbZEOQKv (ORCPT ); Fri, 15 May 2009 12:10:51 -0400 Subject: RE: v4l-dvb rev 11757 broke building under Ubuntu Hardy From: Andy Walls To: chaithrika Cc: david.ward@gatech.edu, linux-media@vger.kernel.org In-Reply-To: <027f01c9d55d$3beea460$b3cbed20$@com> References: <1242345230.3169.49.camel@palomino.walls.org> <027f01c9d55d$3beea460$b3cbed20$@com> Content-Type: text/plain Date: Fri, 15 May 2009 12:08:39 -0400 Message-Id: <1242403719.26764.22.camel@morgan.walls.org> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Sender: linux-media-owner@vger.kernel.org List-ID: On Fri, 2009-05-15 at 18:31 +0530, chaithrika wrote: > Andy, > > Thanks for the patch. Please see my response below. > > > -----Original Message----- > > From: Andy Walls [mailto:awalls@radix.net] > > Sent: Friday, May 15, 2009 5:24 AM > > To: david.ward@gatech.edu > > Cc: Chaithrika U S; linux-media@vger.kernel.org > > Subject: Re: v4l-dvb rev 11757 broke building under Ubuntu Hardy > > > > > > David Ward wrote: > > > > > I am using v4l-dvb in order to add the cx18 driver under Ubuntu Hardy > > > (8.04). > > > > > > The build is currently broken under Hardy, which uses kernel 2.6.24. > > I > > > have traced the origin of the problem to revision 11757. As seen in > > > the latest cron job output, the build produces the error when trying > > > to compile adv7343.c: > > > > > > /usr/local/src/v4l-dvb/v4l/adv7343.c:506: error: array type has > > incomplete element type > > > /usr/local/src/v4l-dvb/v4l/adv7343.c:518: warning: initialization > > from incompatible pointer type > > > /usr/local/src/v4l-dvb/v4l/adv7343.c:520: error: unknown field > > 'id_table' specified in initializer > > > > > > Thanks for resolving this. > > > > > > David Ward > > > > David, > > > > Please try the patch below. > > > > Chaithrika, > > > > Please review (and test if it is OK) the patch below. It modifies > > adv7343.c to what the cs5345.c file does for backward compatability. > > > > It adds some checks against kernel version, which would not go into the > > actual kernel, and changes some code to use the v4l2 i2c module > > template > > from v4l2-i2c-drv.h, which *would* go into the actual kenrel. > > > > > > Regards, > > Andy > > > > > > Signed-off-by: Andy Walls > > > > diff -r 0018ed9bbca3 linux/drivers/media/video/adv7343.c > > --- a/linux/drivers/media/video/adv7343.c Tue May 12 16:13:13 2009 > > +0000 > > +++ b/linux/drivers/media/video/adv7343.c Thu May 14 19:51:10 2009 - > > 0400 > > @@ -29,6 +29,8 @@ > > #include > > #include > > #include > > +#include > > +#include "compat.h" > > > > #include "adv7343_regs.h" > > > > @@ -503,6 +505,7 @@ > > return 0; > > } > > > > +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 26) > > static const struct i2c_device_id adv7343_id[] = { > > {"adv7343", 0}, > > {}, > > @@ -510,25 +513,12 @@ > > > > MODULE_DEVICE_TABLE(i2c, adv7343_id); > > > > -static struct i2c_driver adv7343_driver = { > > - .driver = { > > - .owner = THIS_MODULE, > > - .name = "adv7343", > > - }, > > +#endif > > +static struct v4l2_i2c_driver_data v4l2_i2c_data = { > > + .name = "adv7343", > > .probe = adv7343_probe, > > .remove = adv7343_remove, > > +#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 26) > > .id_table = adv7343_id, > > +#endif > > }; > > - > > -static __init int init_adv7343(void) > > -{ > > - return i2c_add_driver(&adv7343_driver); > > -} > > - > > -static __exit void exit_adv7343(void) > > -{ > > - i2c_del_driver(&adv7343_driver); > > -} > > - > > -module_init(init_adv7343); > > -module_exit(exit_adv7343); > > > > During my initial implementations, I did follow the v4l2_i2c_driver_data > procedure. > But after the review comment I got, I changed it. Please refer to the link > below. > > http://www.mail-archive.com/linux-media@vger.kernel.org/msg02973.html Oh, OK. What prompted me to suggest that was the adv7343_probe() function had the wrong prototype (the incompatible pointer type warning) for that 2.6.24 kernel. The v4l2_i2c_driver_data method was the easiest way around it. Perhaps it would just be easier to add the adv7343 driver to the [2.6.26] section of v4l/versions.txt and forget about backward compatibility for this module. > Was wondering if this is related to kernel version and is there a > possibility that in the > above patch, the module init and exit functions can be included in kernel > version checks > instead of removing them altogether. Well if Hans said v4l2-i2c-drv.h is going away, it's best to avoid trying to use it. If you want to provide backward compatibility for kernels older than 2.6.26, then the 2 version checks that I put in are needed. You'll need to figure out the rest of the i2c structure differences and wrap those in kernel version checks too. If you don't want to provide backward compatibility, then you'll need to add an entry in v4l/versions.txt so that builds on older kernels won't try to compile the module. Regards, Andy > Regards, > Chaithrika