| Age | Commit message (Collapse) | Author |
|---|
|
libelf is deprecated and does not build with gcc-14.
Use elfutils instead as drop-in replacement.
Add patches from Buildroot to disable progs, which use glibc
specific functions.
|
|
Wrong ordering in calloc() is bad, mkay?
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
The 'bounded' attribute seems to never have existed. Replace it by the
'access' attribute which probably does the same. Declare decompressors'
'dst' parameter as read_write for now, no idea if any of them reads back
the written output while decompressing.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
The asprintf() function is not exported unless _GNU_SOURCE has been
defined.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
The callback declaration adjustment was flawed: While dropping
'restrict' keyword might be needed for some reason, dropping the
asterisk changes parameter types. Also, 'const data' defaults to 'int'
type when it should be 'void' instead.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
Kudos to Fabrice Fontaine for coming up with this:
https://lore.kernel.org/buildroot/20240113100654.85904-1-fontaine.fabrice@gmail.com/
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
This reverts commit c5dc60b664fcbba2642346c4b80fc407264337b7.
This change breaks when compiling using gcc-14. It is also not correct:
on_exit() takes two arguments and passes the exit status to the callback
as expected by trace_on_exit() here. In contrast, atexit() expects a
callback which does not take any arguments and thus the logic for when
to print a trace breaks when using it.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
|
|
Without it, on these architectures 'lvm pvscan' will busy-loop printing:
| io_getevents failed: No such file or directory
The symbol is defined in kernel's arch/Kconfig with 'default !64BIT ||
COMPAT'. Introduce Config.in.compat which has the symbol default to y
unless a known 64bit ADK_TARGET_ARCH is enabled. Drop the symbol from
all target/*/kernel/* for which the default applies.
On 64bit systems, enabling this symbol is probably useful to provide
comaptibility towards 32bit user space. Not sure if this is a possible
choice in OpenADK, but just in case the symbol was not removed from:
- target/aarch64/kernel/raspberry-pi4-64
- target/mips64/kernel/qemu-mips64
- target/riscv64/kernel/qemu-riscv64
Another case is targets explicitly having the symbol disabled, namely
kvx. Given that the two kernel configs in question are full copies of
the kernel-generated one, they are likely not a deliberate choice and
thus removed as well.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
This is a more capable alternative to gmediaserver.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
This still lacks some init scripts to reassemble pools and for other
tasks. For now, I just do:
| disks=$(blkid | sed -n 's,^/dev/\(sd[a-z]\): .* TYPE="crypto_LUKS".*,\1,p')
| for disk in $disks; do
| [ -b /dev/mapper/$disk ] || cryptsetup luksOpen \
| --key-file=/etc/mykeyfile /dev/$disk $disk
| done
| zpool import -a
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
For USB-connected UPSs, kernel's USB_HIDDEV option is required as the
daemon communicates via /dev/usb/hiddev* raw HID devices. Therefore make
USB_HID select USB_HIDDEV (shouldn't hurt much on other systems) and
select the former if the package is enabled.
Ship a config which defaults to USB-based UPSs as they are probably the
most common ones. Also install a minimal apccontrol which merely
remounts relevant filesystems read-only in case of near blackout. The
core OpenADK fs layout should tolerate pulling the plug as-is already.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
Needed by upcoming minidlna package.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
Also add a sample LACP configuration and while being at it, fix the
documentation URL.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
Building man pages requires pod2man tool. Avoid this extra build
dependency.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
This actually works around a problem when building against uClibc which
does not seem to provide process_vm_readv syscall and thus compiling
lsfd fails.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
Python is usually installed in /usr/bin.
Actually this should use host-built python binary instead, but porting
the dependencies seems non-trivial. Python packaging is such a mess. :(
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
The included mk/python.mk was removed by commit 179b687e2f99d ("remove
python2 dependencies and packages"). Including python3.mk might be the
right alternative?
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
The old one didn't compile for me anymore, probably due to gcc-12.4.0
being more picky than the one used previously.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
This is required, make sure things won't break if a config is possible
which does not already pull it in via some other dependency.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
Just like with libffi package, HOST_STYLE=auto led to 'prefix=/usr' in
host build libffi.pc.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
This would also contain 'prefix=/usr'. Not sure what this 'auto'
HOST_STYLE is supposed to do, but in this case it breaks things.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
The file contained 'prefix=/usr' which led to glib host build trying to
link against the host's libz.so instead of the built one.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
They moved in commit 724ba67515320 ("ARM: dts: Move .dts files to vendor
sub-directories"), try to cover for it.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
This file is present in ffmpeg sources since version 2.1 already and due
to internal changes, it wont link anymore (undefined reference to
ff_imdct_half_c). Just drop it.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
There is a package choice for libffmpeg to select the SSL backend
library. It was buggy though, the build dependency was not recognized.
Looks like the choice has to use PKG_LIBNAME to be recognized by
depmaker.
Fixes: b076cd3c45c85 ("ffmpeg: add choice for ssl")
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
Configure script incorrectly "detected" no flags are necessary. The odd
result is that 'ls <largefile>' prints huge file size (16E) for files
larger than 2GB while 'du <largefile>' is correct in mounted shares.
Signed-off-by: Phil Sutter <phil@nwl.cc>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
