SliTaz Forum
You are not logged in.
>aufs patchset
One idea for some distant future - maybe look at overlayfs, that's included in the kernel with no external patches.
Offline
Hi devl547,
For aufs you are right, it's on my todo to switch on overlayfs 
.
At time, all stuff for cook based on it and implied rework cookutils, why i keep it for moment.
I launch kernel build, (skip x86_64 for moment, for 4.x, we need cross toolchain, issue with uclibc prebuilt).
Offline
build kernel 4.19.294 ok, but need tweak receipts for modules.
Infact few now included (acpi/video acpi/button, ...), other become mainline instead staging (usbip, rt8192ee, ...)
Offline
@shann
>slitaz-current-20230903.iso work as expected
Works correctly, no problem booting.
Offline
Offline
@devl547
I put new iso with syslinux build, i see that forget to set iso as hybrid .
Now it's ok, new isos pushed https://people.slitaz.org/~shann/slitaz-current-stuff/new_isos/
Offline
For kernel, i relaunch build,
in case of modules i see :
linux-wireless, r8192ee and rtl8821ae now mainline (crystalhd removed)
linux-kvm, lguest module removed since 4.14
for linux-sound, after build linux, seem missing firmware folder.
I go to check and adjust next receipt for firmware, i used linux config that i bump to 4.19 (come from 4.14)
Offline
Hi,
Yesterday i continue adjust kernel config to have all modules need, and adjust modules.list with right path of modules (staging to mainline, subpath for usb udc device, etc.)
also adjust usbip receipt (staging to mainline), and tweak linux receipt to avoid try to cook linux64 modules if "--nox64" is set.
By default no condition set and try build all wanted linux including linux64-*.
Offline
>Now it's ok, new isos pushed
Checked, core iso works correctly. But the size >_>
Looks like we need to do something with libwebkit and all other big stuff. Waiting to get my hands dirty)
Offline
For size, i check with slitaz-rolling-5in1.
Decompressed in ram same size ~230M, diff i see compress rootfs.gz, in my case i used gzip -9 to repack rootfs.gz when i add packages midori/webkit, rolling use lzma.
Offline
After tweaks kernel config to ensure we no miss modules, we have now linux-4.19.294 and linux-* modules.
And also aufs stuff for 4.19 (aufs / aufs-utils).
Need check why we have 171 modules are orphans 
(in case i think that due of change 3.x to 4.x)
Offline
I check for cooking kernel 3.16.55, same above, we have orphans modules.
In my case 4.19.294 => 170 orphans vs 150 for 3.16.55 (after clean, sort -u unpacked modules file, we have both 3.16.55-slitaz/3.16.55-slitaz64)
i go to check diff, but in case seem normal
Offline
Diff i see from 3.16.55 / 4.19.294
ad7291.ko.xz (linux-libre)
bcm-phy-lib.ko.xz (only 4.19)
cfg80211.ko.xz (need check normally in linux-wireless)
charlcd.ko.xz (only 4.19)
cpuid.ko.xz (linux-libre)
dwc3-haps.ko.xz (only 4.19)
et131x.ko.xz (linux-libre / linux-staging)
intel_soc_dts_iosf.ko.xz (only 4.19)
isl29018.ko.xz (linux-libre / linux-staging)
libahci_platform.ko.xz (linux-libre)
libcomposite.ko.xz (linux-libre)
mpls_gso.ko.xz (linux-libre)
mtd.ko.xz (linux-libre)
nf_defrag_ipv4.ko.xz (need check normally in linux-netfilter)
nf_defrag_ipv6.ko.xz (need check normally in linux-ipv6)
nf_reject_ipv4.ko.xz (only 4.19)
nsh.ko.xz (only 4.19)
panel.ko.xz (linux-libre / linux-staging)
pn533_i2c.ko.xz (only 4.19)
ptp_kvm.ko.xz (only 4.19)
rmi_core.ko.xz (only 4.19)
rmi_i2c.ko.xz (only 4.19)
sctp_diag.ko.xz
slicoss.ko.xz (linux-libre / linux-staging)
snps_udc_core.ko.xz (only 4.19)
st21nfca_hci.ko.xz (linux-libre)
tsl2583.ko.xz (linux-libre / linux-staging)
uio.ko.xz (linux-libre)
usb_f_mass_storage.ko.xz (linux-libre)
vhost.ko.xz (linux-libre)
virtio_console.ko.xz (only 4.19)
vport-gre.ko.xz (only 4.19)
vsock_diag.ko.xz (only 4.19)
xhci-plat-hcd.ko.xz (linux-libre)
Except for this below i need check why no push in package, other is normal
[c]
cfg80211.ko.xz (need check normally in linux-wireless)
nf_defrag_ipv4.ko.xz (need check normally in linux-netfilter)
nf_defrag_ipv6.ko.xz (need check normally in linux-ipv6)
[/c]
Concerning (Only in 4.19), need check if we can remove it or if new depends.
Offline
Hi again
i adjust linux-netfilter / linux-ipv6 and linux-wireless receipt to have workaround for copy miss modules.
Also create new packages for linux-ubifs / linux-mtd (used for embedded world) and linux-openvswitch.
Now we have ~90 orphans, continue to check for dispatch modules can be.
Offline
Hi,
Used SliTaz for do labs in my embedded training is really nice.
Infact, we need generate system with buildroot, try to build it with toolchain, catch two issues :
In file included from libfakeroot_time64.c:13:
/usr/include/fts.h:41:3: error: #error "<fts.h> cannot be used with -D_FILE_OFFSET_BITS==64"
# error "<fts.h> cannot be used with -D_FILE_OFFSET_BITS==64"
due of old glibc 2.21, after bump to 2.23 work nice.
But again issue when host-cmake build, for unknown reason we catch race condition.
I think about gcc 6.3.0 and libatomic, i bump to 8.3.0, but issue remain.
undefined reference to `__atomic_fetch_add_8'
undefined reference to `__atomic_store_8'
After search seem already report one month ago, https://gitlab.kitware.com/cmake/cmake/-/issues/25114
In case, i think we need switch to glibc 2.23 (backward compat with 2.21), but of course to sanity, need rebuild.
And also switch to gcc 8.3.0.
Offline
Night good advice.
This catch for really new recent apps (cmake 3.27.x).
For glibc, i think we can bump it, but for gcc not sure.
In case of libc, most apps have issue with big jump between 2.14.1 > 2.21, with this normally can be smooth.
I continue to check for libwebkit, try to build webkitgtk 1.10, (few patch from gentoo to avoid break build with ICU 70), build in progress.
In same time need to create cross toolchain (binutils / gcc) for x86_64 and build 64bit version of 4.19.294 (at time only x32).
Offline
webkitgtk break build with issue in Source/WebCore/dom/Document.cpp and Source/WebCore/platform/graphics/SegmentedFontData.cpp.
miss ';' expected
Found patchs (thanks Peter) http://lists.busybox.net/pipermail/buildroot/2020-February/575223.html
build launch again.
Offline
always in build for webkit.
patch for fix build with icu
patch for fix expected ';'
patch for fix mix boolean in JPEGImagedecoder
and last patch for build with gcc6 / glibc 2.23 (hope is last i need apply)
Offline
Build failed with :
collect2: error: ld returned 1 exit status
/usr/bin/ld: ./.libs/libwebkitgtk-1.0.so: undefined reference to [c]webkit_dom_dom_security_policy_get_type[/c]
/usr/bin/ld: ./.libs/libwebkitgtk-1.0.so: undefined reference to [c]webkit_dom_storage_info_get_type[/c]
/usr/bin/ld: ./.libs/libwebkitgtk-1.0.so: undefined reference to [c]webkit_dom_geolocalisation_get_type[/c]
/usr/bin/ld: ./.libs/libwebkitgtk-1.0.so: undefined reference to [c]WebCore::openGLFunctionTable()[/c]
/usr/bin/ld: ./.libs/libwebkitgtk-1.0.so: undefined reference to [c]WebCore::initializeOpenGLShims()[/c]
I use make -j4 after set mesa-dev dep to accelerate build, not sure related but fallback to make -j1 default
Offline
I see this bug seem related https://bugs.gentoo.org/458524.
In case we use --disable-gl, but during build, complain about gl.h ??
I change flag to --enable-gl, build launch again.
Offline
I retry build libwebkit but same issue.
Quick workaround to test build without libstdc++ gcc83 issue, i set [c]find / -name '*.la' -delete[/c] to remove .la file at begin of receipt.
But same issue with make -j1
collect2: error: ld returned 1 exit status
/usr/bin/ld: ./.libs/libwebkitgtk-1.0.so: undefined reference to webkit_dom_dom_security_policy_get_type
/usr/bin/ld: ./.libs/libwebkitgtk-1.0.so: undefined reference to webkit_dom_storage_info_get_type
/usr/bin/ld: ./.libs/libwebkitgtk-1.0.so: undefined reference to webkit_dom_geolocalisation_get_type
Offline
[ Generated in 0.019 seconds, 7 queries executed - Memory usage: 1.59 MiB (Peak: 1.77 MiB) ]