Re: [RFC PATCH] media_build: two fixes + one unresolved issue

[Mauro Carvalho Chehab <mchehab@redhat.com>]

Em 05-10-2011 06:23, Hans Verkuil escreveu:
> Hi Mauro,
> 
> While doing a compatibility build I found three issues. I've got patches for two,
> but one issue is still unresolved.
> 
> The first is this small patch to get rid of this warning when doing 'make install':
> 
> make -C firmware install
> make[2]: Entering directory `/home/hve/work/media_build/v4l/firmware'
> Installing firmwares at /lib/firmware: vicam/firmware.fw dabusb/firmware.fw dabusb/bitstream.bin ttusb-budget/dspbootcode.bin 
> cpia2/stv0672_vp4.bin av7110/bootcode.bin *.fw* cp: target `/lib/firmware/v4l-pvrusb2-29xxx-01.fw' is not a directory
> make[2]: [install] Error 1 (ignored)
> 
> The fix is simply to remove '*.fw*' since it doesn't match any files.
> 
> diff --git a/v4l/firmware/Makefile b/v4l/firmware/Makefile
> index fb53ef2..bcbc784 100644
> --- a/v4l/firmware/Makefile
> +++ b/v4l/firmware/Makefile
> @@ -22,7 +22,7 @@ distclean: clean
>  install: default
>  	@echo -n "Installing firmwares at $(FW_DIR): "
>  	-@for i in $(DIRS); do if [ ! -d $(FW_DIR)/$$i ]; then mkdir -p $(FW_DIR)/$$i; fi; done
> -	-@for i in $(TARGETS) *.fw*; do echo -n "$$i "; cp $$i $(FW_DIR)/$$i; done
> +	-@for i in $(TARGETS); do echo -n "$$i "; cp $$i $(FW_DIR)/$$i; done
>  	@echo

I suspect that, for some unknown reason, you're not capable of downloading the firmwares from
linuxtv.org, or maybe you never ran the build script. The build firmware downloads the latest
version of the firmwares from:
	http://www.linuxtv.org/downloads/firmware/

It should be noticed that the install procedure won't fail if you never downloaded the firmwares,
due to the "-" signal. So, please don't apply this patch.

>  rminstall:
> 
> I think this fix is fine, unless this is something you want to have for the future.
> 
> The other is a kernel naming issue: my aptosid distro (debian based) running
> kernel v3.0.0 uses a different naming convention:
> 
> $ uname -r
> 3.0-4.slh.6-aptosid-amd64
> 
> So the sublevel is not shown.
> 
> This patch makes the sublevel optional (and assumes it to be 0 if absent):
> 
> diff --git a/linux/patches_for_kernel.pl b/linux/patches_for_kernel.pl
> index c19b216..33348d9 100755
> --- a/linux/patches_for_kernel.pl
> +++ b/linux/patches_for_kernel.pl
> @@ -13,8 +13,11 @@ my $file = "../backports/backports.txt";
>  open IN, $file or die "can't find $file\n";
>  
>  sub kernel_version($) {
> -	$_[0] =~ m/^(\d+)\.(\d+)\.(\d+)/;
> -	return ($1*65536 + $2*256 + $3);
> +	my $sublevel;
> +
> +	$_[0] =~ m/^(\d+)\.(\d+)\.?(\d*)/;
> +	$sublevel = $3 == "" ? 0 : $3;
> +	return ($1*65536 + $2*256 + $sublevel);
>  }
>  
>  my $kernel = kernel_version($version);
> diff --git a/v4l/Makefile b/v4l/Makefile
> index ab07a7a..311924e 100644
> --- a/v4l/Makefile
> +++ b/v4l/Makefile
> @@ -248,7 +248,7 @@ ifneq ($(VER),)
>  	@echo $(VER)|perl -ne 'if (/^([0-9]*)\.([0-9])*\.([0-9]*)(.*)$$/) { printf 
> ("VERSION=%s\nPATCHLEVEL:=%s\nSUBLEVEL:=%s\nKERNELRELEASE:=%s.%s.%s%s\n",$$1,$$2,$$3,$$1,$$2,$$3,$$4); };' > $(obj)/.version
>  else
>  	@echo No version yet, using `uname -r`
> -	@uname -r|perl -ne 'if (/^([0-9]*)\.([0-9])*\.([0-9]*)(.*)$$/) { printf ("VERSION=%s\nPATCHLEVEL:=%s\nSUBLEVEL:=%s\nKERNELRELEASE:=%s.
> %s.%s%s\n",$$1,$$2,$$3,$$1,$$2,$$3,$$4); };' > $(obj)/.version
> +	@uname -r|perl -ne 'if (/^([0-9]*)\.([0-9])*\.?([0-9]*)(.*)$$/) { printf 
> ("VERSION=%s\nPATCHLEVEL:=%s\nSUBLEVEL:=%s\nKERNELRELEASE:=%s",$$1,$$2,$$3==""?"0":$$3,$$_); };' > $(obj)/.version
>  endif
>  endif

Seems OK to me. If you're willing to fix those distro-specific stuff, on Fedora 15, 
the 3.0 kernel were renamed as "2.40", as they wanted to avoid touching on some scripts
(that's what I got from a lwn discussion). Maybe other distros might have done weird
things like that. The only practical consequence I noticed with F15 is that one driver
were disabled (the firewire one).

> The last issue I have is that the media.ko module isn't installed when I run
> 'make install'. I tried to fix it, but I got lost in the Makefile/perl magic :-)

That's weird... I took a look here. My Makefile.media was generated with a line
to install it. See:

@n=0;for i in dvb-ttpci.ko budget-patch.ko ttpci-eeprom.ko budget-av.ko budget.ko budget-core.ko budget-ci.ko;do if [ -f 
"$$i" ]; then if [ $$n -eq 0 ]; then echo -n "	dvb/ttpci/: "; install -d $(DESTDIR)$(KDIR26)/dvb/ttpci; fi; n=$$(($$n+1)); if [ 
 $$n -eq 4 ]; then echo; echo -n "		"; n=1; fi; echo -n "$$i "; install -m 644 -c $$i $(DESTDIR)$(KDIR26)/dvb/ttpci; 
fi; done; if [  $$n -ne 0 ]; then echo; strip --strip-debug $(DESTDIR)$(KDIR26)/dvb/ttpci/*.ko; fi;


@n=0;for i in media.ko;do if [ -f "$$i" ]; then if [ $$n -eq 0 ]; then echo -n "	../linux/drivers/media/: "; insta
ll -d $(DESTDIR)$(KDIR26)/../linux/drivers/media; fi; n=$$(($$n+1)); if [  $$n -eq 4 ]; then echo; echo -n "		"; n=1; f
i; echo -n "$$i "; install -m 644 -c $$i $(DESTDIR)$(KDIR26)/../linux/drivers/media; fi; done; if [  $$n -ne 0 ]; then echo; stri
p --strip-debug $(DESTDIR)$(KDIR26)/../linux/drivers/media/*.ko; fi;

For me, it is likely a trouble at scripts/make_makefile.pl, on that line of
the script:
		$idir =~ s|^../linux/drivers/media/||;

An one line patch to make the last / optional is probably enough to fix it.

Care to write the patch and fix it?
> 
> Let me know if you are OK with these two patches, and if you can fix the last
> issue (or give me some hints on how to fix it), then that would be great!

Thanks,
Mauro