| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
| |
The guix style command reformats a package or a file to use the code
style used by Guix.
This commits sync the %commit definition with the output
produced by guix style.
This commit should contain no functional changes.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The guix style command reformats a package or a file to use the code
style used by Guix.
This commits sync the %local-source definition with the output
produced by guix style.
This commit should contain no functional changes.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The guix style command reformats a package or a file to use the code
style used by Guix.
This commits sync the android-make definition with the output produced
by guix style.
This commit should contain no functional changes.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The guix style command reformats a package or a file to use the code
style used by Guix.
This commits sync the android-phases definition with the output
produced by guix style.
This commit should contain no functional changes.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The guix style command reformats a package or a file to use the code
style used by Guix.
This commits sync the libsamsung-ipc package with the output produced
by guix style.
This commit should contain no functional changes.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The guix style command reformats a package or a file to use the code
style used by Guix.
This commits sync the libsamsung-ipc-gcc-android package with the
output produced by guix style.
This commit should contain no functional changes.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The guix style command reformats a package or a file to use the code
style used by Guix.
This commits sync the libsamsung-ipc-gcc-autotools package with the
output produced by guix style.
This commit should contain no functional changes.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The guix style command reformats a package or a file to use the code
style used by Guix.
This commits sync the libsamsung-ipc-clang-autotools package with the
output produced by guix style.
This commit should contain no functional changes.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The guix style command reformats a package or a file to use the code
style used by Guix.
This commits sync the libsamsung-ipc-clang-android package with the
output produced by guix style.
This commit should contain no functional changes.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The guix style command reformats a package or a file to use the code
style used by Guix.
This commits sync the package imports with the output produced by guix
style.
This commit should contain no functional changes.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
| |
The https-send-sms tool is linked to libcurl, so libcurl is needed at
runtime.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
| |
This commit should contain no functional changes.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Using python instead of python3 has several issues:
- Distributions and users can choose python2 as their default python
interpreter. For users this can be done by making a symlink in
/usr/local for instance.
- Guix doesn't have python but has python3 in the PATH. Changing to
python3 makes tests work for Guix system users that use the
scripts/manifest.scm file with guix shell / guix environment.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
| |
Tabs were added by the commit 43983cdb532303224be896ec37e67fa515d87e3e
("Android.mk: fix missing files and build local-modules in guix.scm").
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
| |
This should help us catch more issues as the PKGBUILD doesn't
necessarily use the same compiler versions than in the guix.scm.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In the commit 5f97bb60091a1c4d4335fd02eff2b01e27e7a881 ("tools:
ipc-modem: Move log functions in their own file"), we added new files
for ipc-modem.
That commit was then tested with scripts/guix.scm, which at the time
didn't build ipc-modem, so the issue slipped through the automatic
tests.
So in addition of adding back these missing files in Android.mk, this
commit also makes guix.scm actually build all the local modules.
This was not done before as it required some extensive work.
That work consisted in adding code to retrieve the LOCAL_MODULEs from
the Android.mk, and passing that to the package build and install
procedures (which had to be converted into Guix's G-Expressions
(gexp)[1] to be able to more cleanly get that data). In addition,
making contributions to android-make-stub was needed as well.
Having some of the parsing done in the guix.scm file instead of
android-make-stub is needed because otherwise the android-make-stub
design would probably need to be redone: android-make-stub consists of
some Makefiles that are passed to make via -f, and that in turn
these Makefiles include the Android.mk in the local directory.
This doesn't make it possible to handle dependencies cleanly, as to do
that, the Android.mk would need to first be parsed through another
mean to understand the dependencies tree, and then to build the local
modules in order.
And that would need to be done in some compatible way not to break
existing users of android-make-stub (like Guix that builds some
Android packages with it).
[1]https://guix.gnu.org/en/manual/devel/en/guix.html#G_002dExpressions
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The CFLAGS that are in the %common-strict-cflags in scripts/guix.scm
have been validated with both GCC and CLANG and they are useful to
find potential issues in the code or code that won't compile on
Android.
However while the scripts/guix.scm script is really useful to test
commits once they are ready, it is less convenient to use it when
fixing issues in commits that are still being worked on.
This is because it is not as fast as building libsamsung-ipc directly
because:
- it requires a clean source directory to work (this can be done with
make distclean) so we can't reuse the current build output
- libsamsung-ipc source code is copied and built 5 times (in different
configurations)
As for the implementation, AM_CFLAGS was used instead of appending to
the default CFLAGS as CFLAGS is meant to be a variable for users. The
effect is that both are independent, so if users don't want strict
CFLAGS, they would need to not use --enable-strict-cflags.
And it was implemented as a shell script to at the same time:
- Enable to have comments, and good formating of the flags.
- Enable to share the cflags between guix.scm and the autotools.
- Keep the complexity low to keep it working in most situations.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This can then be used during automatic tests, but the downside is that
for now this tool only works with one provider.
The following symbols from (lib)curl (along with their corresponding
minimal (lib)curl version) have been used:
+------------------------+---------------------------+
| curl symbols | minimum (lib)curl version |
+------------------------+---------------------------+
| CURL_HTTP_VERSION_2TLS | 7.47.0 |
| CURLUPART_URL | 7.62.0 |
| CURLUPART_QUERY | 7.62.0 |
| CURLU_URLENCODE | 7.62.0 |
| CURLU_APPENDQUERY | 7.62.0 |
| CURLU | 7.62.0 |
| CURLU_ALLOW_SPACE | 7.78.0 |
+------------------------+---------------------------+
So we need to depend on curl 7.78.0.
And we currently have the following curl version:
+----------------------------+--------------------+------------+
| Distribution | (lib)curl version | Compatible |
+----------------------------+--------------------+------------+
| Replicant 4.2 | Doesn't have curl | No |
| Guix 6b9105e557 | 7.28.1 | No |
| Trisquel 7 x86_64 | 7.35.0-1ubuntu2.20 | No |
| Replicant 6 | 7.43.0 | No |
| Debian stretch x86_64 | 7.52.1-5+deb9u10 | No |
| Replicant 11 | 7.67.0 | No |
| Parabola i686 (07/04/2022) | 7.79.0-3.0 | Yes |
| Guix i686 (07/04/2022) | 7.79.1 | Yes |
+----------------------------+--------------------+------------+
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This enables contributors to run Valgrind when running 'make check'
for instance before sending patches.
The --enable-valgrind option was used instead of using
--enable-checking=valgrind like GCC uses as the code for that is much
easier to get right.
As Valgrind is very slow, I had to increase the timeouts in the
ipc-modem tests. For now I settled on 60 seconds to have some margin
for machines or setups that are potentially slower than the one I use.
For the guix.scm, with Guix 1.3.0, Valgrind needs the ld.so .symtab to
work, so as Valgrind doesn't know where to find them, so we had to use
--extra-debuginfo-path for that.
This is also why we pass '-v' to valgrind as it prints message
explaining the issue when it doesn't find the required symbols.
And we also needed a trick mentioned in the Guix mailing list[1] to
find the right path for the debug information as the glibc being used
by packages isn't exported publicly.
[1]https://lists.gnu.org/archive/html/help-guix/2022-03/msg00036.html
The proper fix will be merged in the next Guix release (merging it now
is not possible since it would require to rebuild everything, and that
is only possible with new releases).
A temporary workaround with 4d1a88312cbebb10d47905d6d023953ac85bcd04
(gnu: valgrind: Allow ld.so symbols to be found.) was also merged in
the current Guix so it is possible to get it with guix pull, but here
it's probably better to stick with a workaround in the guix.scm for
now as this also enables to use older guix revisions for testing.
Thanks to the people on #guix on Liberachat for helping me to solve
this Valgrind issue.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
| |
Replicant 6.0 is currently the main project using libsamsung-ipc and
it uses --enable-debug to show the hexadecimal dump of the samsung-ipc
packets, so it's a good idea to do compilation tests with it.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Without the commit 355433ab5f17f25ca202ec092d989b267fd78c19
(tools: nv_data_imei: get rid of missing initializer warning),
we have the following warnings when building Replicant 6.0:
hardware/replicant/libsamsung-ipc/tools/nv_data-imei.c:245:2:
warning: missing initializer for field 'option' of 'struct
command_option' [-Wmissing-field-initializers]
{ /* Sentinel */ },
^
In file included from
hardware/replicant/libsamsung-ipc/tools/nv_data-imei.c:39:0:
hardware/replicant/libsamsung-ipc/tools/nv_data-imei.h:65:10:
note: 'option' declared here
uint8_t option;
^
hardware/replicant/libsamsung-ipc/tools/nv_data-imei.c:277:2:
warning: missing initializer for field 'name' of 'struct command'
[-Wmissing-field-initializers]
{ /* Sentinel */ },
^
In file included from
hardware/replicant/libsamsung-ipc/tools/nv_data-imei.c:39:0:
hardware/replicant/libsamsung-ipc/tools/nv_data-imei.h:57:14:
note: 'name' declared here
const char *name;
^
The addition of -Werror=pedantic in guix.scm enables to detect these
as the build would then fail:
CC nv_data-imei.o
nv_data-imei.c:245:2: error:
ISO C forbids empty initializer braces [-Werror=pedantic]
245 | { /* Sentinel */ },
| ^
nv_data-imei.c:277:2: error:
ISO C forbids empty initializer braces [-Werror=pedantic]
277 | { /* Sentinel */ },
| ^
CC nv_data-md5.o
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
| |
This makes it easier to build-test all the latest commits before
sending patches or pushing patches.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
| |
The usage of this PKGBUILD is similar to the guix.scm file in the same
directory: it can be used to test the build of the current source.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
On Guix, if we have all the required dependencies installed, running
./autogen.sh && ./configure && make fails with the following error:
CCLD libsamsung-ipc-test
ld: /gnu/store/cm6wv72061h65lri2aqgwa8bvncls51h-openssl-1.1.1l/lib/libcrypto.so: undefined reference to `fstat@GLIBC_2.33'
ld: /gnu/store/cm6wv72061h65lri2aqgwa8bvncls51h-openssl-1.1.1l/lib/libcrypto.so: undefined reference to `stat@GLIBC_2.33'
The same command works when using guix shell or guix environment
(still with all the dependencies).
Having a manifest file in libsamsung-ipc will enable Guix system users
to easily work on libsamsung-ipc without having to use trial and error
to get a working setup.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Libraries typically have a public API and some internal
implementation of that API. In libsamsung-ipc, the public
API is defined in the include/ directory from the top directory.
When compiling and installing libsamsung-ipc, that include/
directory is installed as well.
Anything that is not defined in include/ is not part of that
public API.
However here we need to precisely test functions that are not
part of that public API: The open_android_modem_partition
function being tested here is only used by the herolte device,
and that device doesn't have a battery that is easily replaceable,
so most Replicant contributors will not want to get that device.
As currently all the non static symbols of libsamsung-ipc are
exported and that open_android_modem_partition isn't static,
we can simply link to libsamsung-ipc for now and use its
internal headers to access the functions to test.
If at some point, libsamsung-ipc only exports the symbols defined
in include/ we would need to find other ways to run such tests,
for instance by using test frameworks that take care of that or by
compiling libsamsung-ipc source code into the test programs.
Some of the code of the libsamsung-ipc-test utility was adapted
from code from the nv_data-md5 utility (which is currently in
tools/).
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
The TARGET_ERROR_FLAGS are available from build/core/config.mk
in Replicant 6.
-Werror=non-virtual-dtor was only added to CLANG because it fails
to be recognized with GCC:
cc1: error: ‘-Werror=’ argument ‘-Werror=non-virtual-dtor’
is not valid for C [-Werror]
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
| |
The COMMON_GLOBAL_CFLAGS are available from build/core/config.mk
in Replicant 6.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
| |
This enables to more easily add new flags.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Without that, it will end up using clang-10 while we now have
clang-11.
This will make Guix build clang-10 which is time consuming.
Instead by using the latest clang we can use binary packages
(Guix calls that substitutes) of clang which speeds up the
build of libsamsung-ipc a lot, since most of the time we don't
need to build any dependencies.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The scripts directory already has several scripts that are not
needed to use libsamsung-ipc, but all of them are (or were) useful
to work on patches for libsamsung-ipc.
As this also applies to the guix.scm and that scheme is a scripting
language, it would be more clear to have it moved in the scripts
directory.
In addition it could also make it more simple for distributions
to not ship these files if they aren't interested in having them
and/or if they only want files compatibles with the license
currently used by libsamsung-ipc (GPLv2 or later).
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This checks the files that aren't modified by any of the
commits that are in the revision range.
This is not completely failsafe as checkpatch.pl tend to
check better what's touched by a patch, in the case of
modifications that do not touch C code:
- Checking for scripts/spelling.txt with checkpatch.pl -f
doesn't raise any errors or warnings.
- Checking the patch that adds scripts/spelling.txt
raise a lot of warnings.
However checkpatch.pl seem to find issues in C code regardless
of whether the code is checked as part of individual files
or as part of a patch.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
Reviewed-by: Joonas Kylmälä <joonas.kylmala@iki.fi>
|
| |
|
|
|
|
|
|
|
| |
This adds an option to check if all the repository files
were "converted" by checking if they all have been modified
by a patch in the git revision range given.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
Reviewed-by: Joonas Kylmälä <joonas.kylmala@iki.fi>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When sending a patch for Linux, and running checkpatch.pl,
you typically only need to check the patch with it.
However since we're doing an initial conversion to the
Linux code style, we also need to check all the content of
the files as well, as otherwise the parts that are not
being touched by the patches won't be checked at all.
We also need to get it right, as otherwise running
checkpatch.pl on subsequent patches could trigger some
errors that would not come from the patches being checked,
but that would instead come from previous code style
errors, that were left unfixed.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
Reviewed-by: Joonas Kylmälä <joonas.kylmala@iki.fi>
|
|
|
This will be used to check if the patches match Linux code style.
The configuration is needed because otherwise the this tool
has defaults parameters which are made for Linux.
Signed-off-by: Denis 'GNUtoo' Carikli <GNUtoo@cyberdimension.org>
Reviewed-by: Joonas Kylmälä <joonas.kylmala@iki.fi>
|
