Timing detail reporting can be enabled by setting
TRACE_BUILD_TIMING=1
This enables timestamping collecting at various build stages so we
can easily analyze how long eg configure, make/build, install etc
steps take.
Signed-off-by: Matthias Reichl <hias@horus.com>
add flag "-o" to overwrite file on unzipping.
when a build is interrupted by some issues, it will cause a re-substitution problem after a rebuild the target whose source code come from a zip file (like target jzintv)
The buildsystem's Makefile hardcodes the location of the BUILD_DIR used
by the rest of the buildsystem. Add a script to handle what BUILD_DIR
might be set to.
Signed-off-by: Ian Leonard <antonlacon@gmail.com>
* Downgrade glibc to 2.27 for L4T Builds.
* Downgrade binutils to 2.32 for L4T Builds.
* Update L4T Project Options
* Refactor tegra-bsp package
* Import/Rebase CTC's ffmpeg patches, and switch decoder to nvv4l2 from nvmpi. Fixes decoder issues
* Enable all disabled cores
* Remove hack in busybox package to dump firmware into L4T Builds, we now handle this in tegra-bsp package.
* More L4T project cleanups
* Bring switch-bootloader package.mk inline with upstream formatting
* Fix reboot_payload
* Add proper tegra-bsp:init support
* Bootloader cleanups, seperate bootloader stuff for libreelec builds
* Update init for supporting all distros on Nintendo Switch
* Add proper splash images for libreelec builds.
* Final bootloader fixes
* switch-bootloader-merge-fixes
* Fix uboot merge issues
* fix bmp
* It all builds
* add Basic USB Gadget(Serial console only)
* Properly fix mtp, and finish gadget stuff
* grub: update to latest release
* gnulib: update to match grub bootstrap.conf
* grub: build 32 and 64 bit EFI images
* syslinux: do not copy EFI image
* installer: adapt script to above changes
* mkimage: virtual: build only for 64-bit Generic
* mkimage: grub: create cfg / copy EFI images only for 64-bit Generic
* create temporary folder in the target folder
* move DISK_LABEL setting closer to the part where it is used
* te MBR without test (it should be already present and if not, dd
will fail anyway and error is caught)
* update output when vmdk is created
In case SYSTEM_SIZE was higher than 1024, mcopy would fail without the
failure being caught. The failure was probably due to invalid geometry
(number of sectors per track too high?). However mformat should be able
to determine the geometry on its own, when no geometry parameters are
supplied. After discussion on IRC with vpeter we came to this solution.
* create temporary folder in the target folder
* use sparse files to create image and partitions
* no extractions of the partitions from image
* modify mcopy alias to copy to the partition sparse file
* add `-o` switch to mcopy alias, so mcopy does not prompt in case of
existing file in the partition
* remove alias for mformat (is used only one time)
* catch errors for mformat/mcopy/mmd
* use mcopy alias when copying files to virtual appliance part1
* move DISK_LABEL setting closer to the part where it is used
* write MBR without test (it should be already present and if not, dd
will fail anyway and error is caught)
* update output when vmdk is created
Co-authored-by: Peter <peter.vicman@gmail.com>
In case SYSTEM_SIZE was higher than 1024, mcopy would fail without the
failure being caught. The failure was probably due to invalid geometry
(number of sectors per track too high?). After discussion on IRC with
vpeter we came to this solution.
* extract the FAT32 partition from the image
* format the extracted partition without any geometry parameters
(mformat is able to determine the geometry by itself)
* install/copy files to the extracted partition
* verify the partition
* merge the partition back to the image
The extraction of the FAT32 partition is also needed for the OVA image
in case of Generic project.
Errors while mformat/mcopy/mmd will now be caught and image creation
will be terminated with an error message.
Removed alias for mformat, as it is used only once. Added `-o` to alias
for mcopy, just in case, as the output is now redirected, in case a file
would already exist, the prompt would be not shown and image creation
would just hang (ctrl-c does not work, had to use kill -9 on mcopy
during my testing when I forgot the `-o` switch).
For OVA full path to toolchain is not needed, as the alias has already
the partition in it, not the image.
preserves space, tarballs can be reused if the package repository gets
nuked or the commit disappears from the history (e.g. in case of a
forced pushed)
Script uses arbitrary values for FAT32 partition and name of the
distribution. Using placeholders in the script and replacing them at
build time offers flexibility, i.e. no need to change arbitrary values
in script when SYSTEM_SIZE is changed.
Script uses arbitrary values for FAT32 partition and name of the
distribution. Using placeholders in the script and replacing them at
build time offers flexibility, i.e. no need to change arbitrary values
in script when SYSTEM_SIZE is changed.
Eliminate use of cat and subshells in patching.
The previous code takes the first field of PKG_BUILD, using a space
as the delimiter. PKG_BUILD used to be generated using 'ls' at least
as of before commit ef8e871d3c in 2013.
Today, PKG_BUILD is a set path, without spaces.
Signed-off-by: Ian Leonard <antonlacon@gmail.com>
