Age | Commit message (Collapse) | Author |
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
The code to get the file list from hg manifest wasn't escaped right and
the error value check was the opposite of what it should have been.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
aci.c needs some OSS include files from the kernel to compile. Typical
kernel-headers packages don't have these include files, and so aci.c
fails to compile. Have make_kconfig.pl check for the headers, and
disable aci if they aren't there.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
Skip the parsing code that only make sense inside a Kconfig menu block
when not in a menu block. Mostly to get rid of Perl warnings.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
Since Kconfig.sound is processed by make_kconfig.pl like all the other
Kconfig files, it was getting put into v4l/Kconfig twice. Once from a
source directive v4l/Kconfig and again when it was processed by the
script.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
A subtle bug in the code that changed Kconfig depends expressions into
Perl expressions would keep changes to Kconfig variables from
propagating. The subroutine was supposed to change a _copy_ of the
expression from Kconfig to Perl syntax, but was changing the expression
itself.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
Restructure make_kconfig.pl and clean the code up some more.
The dependency tracking code is rewritten. This version converts the
"depends on" and "select X if" lines into Perl expressions, which are
then evaluated. This way all complex expressions are handled correctly.
Dependencies are tracked recursively until all are found. "select" lines
with an "if" clause are handled correctly; the selected variable is only
required when the "if" clause is true.
Another improvement is that hex/string/int variables that don't have
their dependencies met will not appear in the .config file. They aren't
supposed to.
make_kconfig will fail with an error message if MODULES is turned off in
the Kernel's config, as building out of tree makes no sense in that
situation.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
The script which combines the individual Makefils into Makefile.media
would include a Makefile each time it was referenced. This caused some
Makefiles to get included multiple times.
Also print comments in Makefile.media with the names of the Makefiles that
are being included.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
The make_makefile.pl script is re-written. The old one was nasty. This
one will correctly handle Makefile lines that are continued with a
backslash. It will also generate a dependency list of the Makefile used
to create Makefile.media. This way if any change, Makefile.media will be
automatically re-made by make.
The whole business where make_makefile.pl would call make_kconfig.pl is
removed.
The underlying logic of the main Makefile is completely changed. Make
has a feature where it will make sure any included files are up-to-date,
and then restart itself if any were updated. We now take advantage of
this feature.
Detect when make is building only targets which do not need include files
like .myconfig or Makefile.media, and don't include them in those cases.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
Add the ability to the make_kconfig.pl script to create a file
named ".kconfig.dep" which can be included in a Makefile and defines the
dependencies of the v4l/Kconfig file. That is, all the individual
Kconfig files that were used to make it.
Added a rule to v4l/Makefile to remake Kconfig when needed, and include
.kconfig.deps to get Kconfig's dependencies. Adjust the config targets
so that they depend on Kconfig, rather than including the commands to
create Kconfig in the config targets' scripts. The config targets no
longer depend on .version, as they don't need it themselves. Rather,
it's Kconfig that needs .version, so it's made one of Kconfig's
dependencies.
Add Kconfig as a dependency to .myconfig, it should be there since the
make_myconfig.pl script uses Kconfig.
Now if you change any of the Kconfig files, the v4l/Kconfig file will
automatically get remade. This will in turn cause .myconfig to be
re-made, and any new options will appear in .myconfig (turned off).
Ideally this should trigger some kind of oldconfig system to update
.config.
Fix the .version rule so that instead of re-creating .version every time,
it is only created when it didn't already exist. Since Kconfig depends
on .version, it was getting re-created every time too. If .version
existed, it would just get re-created to the same thing anyway.
Stop echoing the echo srcdir command itself and don't echo anything if
SRCDIR is blank.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
New code to read the Kernel's (or v4l-dvb's) '.config' file. This
replaces the code that used to read the config.h/autoconf.h file. In
2.6.18 the autoconf.h file no longer mentions options set to 'n' at all,
while .config has a comment line for each disabled option. Reading
.config is easier and gives us a list of the disabled options.
Restructure the code for reading '.version' too.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Still keeping the warning message for kernel-readers only.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Now, make will compile also the audio drivers at the tree.
By default, they will not be installed. A new option were
added to install those files:
make sound-install
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Those files are not part of V4L stuff. However, they are or dependent of V4L
apis or used on some video boards.
After this patch, it will be possible to compile and test they against changes
at V4L API or drivers.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
"select" were overriding "depends on". A workaround were previously
been applied.
This patch removes the workaround and fixes the issue.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
This script will read a Linux Kernel style Makefile tree and produce an
overview of the build process.
It produces three lists which show to associations of:
1. Kconfig variables to kernel module(s) built
2. Kernel modules to source file(s) that make them
3. Kconfig variables to source file(s) built
It can also check the Makefiles for various errors and warnings.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
This script will check source files against the v4l-dvb and kernel
Kconfig files, to make sure that all CONFIG_XXX defines mentioned in the
source files have corresponding config options in the Kconfig files.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Michael Krufky <mkrufky@linuxtv.org>
This fixes the patch failure when using the "make kernel-links" build method.
We no longer have to patch videodev.h to include compat.h for the in-kernel
build using the v4l-dvb development repository, as the individual card drivers
that need compat.h are already including it.
Signed-off-by: Michael Krufky <mkrufky@linuxtv.org>
|
|
From: Patrick Boettcher <pb@linuxtv.org>
merge from main
Signed-off-by: Patrick Boettcher <pb@linuxtv.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
Normally programs like modinfo are in /sbin, and non-root users often
don't have this in their path. Make the rmmod.pl script look in some
extra directories like /sbin, /usr/sbin, etc. if it can't find the
utilities in the path.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
The building system now auto-generates this dependency, as tristate,
while those options were manually generating it as boolean, generating
a warning.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Patrick Boettcher <pb@linuxtv.org>
merge from main
Signed-off-by: Patrick Boettcher <pb@linuxtv.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
Instead of reading linux/config.h, read linux/autoconf.h to get the
kernel's configuration. The former file has been obsoleted.
In Kconfig 'select' lines, ignore the 'if' clause if one exists. This is
not a permanent solution to correctly parse and evaluate Kconfig files.
That is much harder, this will get building working again now.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Patrick Boettcher <pb@linuxtv.org>
merge: from main
Signed-off-by: Patrick Boettcher <pb@linuxtv.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
The code to compare kernel versions was wrong, it would think 2.7.1 was
older than 2.6.10 because 1 < 10. Fixed.
Changed code to not require all options to be defined in versions.txt.
It was necessary before only because of the way make_myconfig.pl worked,
but that script has changed. Driver sub-options can be omitted from
versions.txt, unless they somehow need a newer kernel than the driver
itself.
In fact, most of the drivers themselves could be removed from
versions.txt, and only put in if there is an actual reason.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
This will allow generating periodic snapshots of the tree.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Patrick Boettcher <pb@linuxtv.org>
sync with master
Signed-off-by: Patrick Boettcher <pb@linuxtv.org>
|
|
From: Hans Verkuil <hverkuil@xs4all.nl>
Signed-off-by: Hans Verkuil <hverkuil@xs4all.nl>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
tda9887 is not really an independent tuner, but part of one. It shouldn't
be ponted at a userlist.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
When the make commit checks for whitespace errors, it will save a copy of
the changes it makes to $(TMP)/v4l_hg_whitespace. The prep_commit_msg.pl
script will look for this file and include a note in the prepared commit
message with diffstat output.
The changes that scripts/cardlist might make are not included, but that
can be added later.
The prep_commit_msg.pl script is effectively totally re-written. This one
will try some extra stuff to some up with the user's e-mail, such as
looking in the ~/.hgrc file.
The Makefile is changed so that temporary files created by make commit
and make whitespace will honor the $TMP environment variable, if set.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
Before:
--- Makefile
+++ - 2006-06-20 22:16:36.320564000 -0700
Now:
--- Makefile
+++ Makefile
patch handles both correctly, but diffstat will use the "+++" filename
rather than the "---" filename, making diffstat output rather useless.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Most V4L/DVB hg tree users are just compiling using kernel headers.
This is bad, due to:
1) the lack of support for make *config
2) the kernel itself have unknown patches from the distros
Anyway, it is useful to provide they some support.
This patch provides a "hack" for such users, by checking if the
sound/oss/aci.h file is present under kernel tree. If not, it will
disable miropcm20 and print a warning about the usage of a kernel-headers
only package.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Building system now looks kernel config.h and autoconf.h files,
seeking for the compiled options at kernel. If something is missing,
it will disable compilation for that driver, printing a warning message.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
There were a circular dependency at scripts/make_makefile.pl that were
affecting mostly building with GNU make 3.81.
Also, scripts/make_noconfig.pl is, in fact, building .myconfig file, so,
better to name it as make_myconfig.pl.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
make 3.81 were suffering some endless loop that might be caused
by Makefile.media rewrites. After this patch, make_makefile will
only write Makefile.media if needed.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
Faster script that doesn't use any temporary files. The old one
would also miss cleaning four spaces in a row in places where
it would clean one to eight spaces. Has two options:
fast
Only clean whitespace in files Hg thinks are modified or added. Used by
make whitespace and much (about 100x) faster.
manifest
Clean all files controlled by Hg, using "hg manifest"
The default with no options is the old behaviour, clean all files under
the linux directory, except CVS files.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
A check of the current Linux 2.6 kernel reveals that the default values
for hex options sometimes have the form "0x1f0" and sometimes "1f0", with
the former being more common. Adjust the make_kconfig.pl script so it
can understand both.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Kernel handling of hex type passes those values prefixed with
0x to the modules.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Trent Piepho <xyzzy@speakeasy.org>
integer/hex/string options were getting omitted from .myconfig, produced
by make_noconfig.pl. They don't need to be there for the Makefile that
includes .myconfig, it only uses the bool/tristate options. They need to
be there for compat-config.h, which is generated from .myconfig and so
will only have the #define statements for options that appear in .myconfig.
Signed-off-by: Trent Piepho <xyzzy@speakeasy.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
rmmod <file> doesn't work with rmmod version 3.2.2. This patch uses
rmmod <module> instead. This should work with older mod-utils.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|
|
From: Mauro Carvalho Chehab <mchehab@infradead.org>
Now, script will detect "hg export" patches inside emails,
removing non-pertinent headers, keeping only the original
commit message and the patch.
Signed-off-by: Mauro Carvalho Chehab <mchehab@infradead.org>
|