Skip to contentSkip to conversation

kernel upgrade breaks image creation

DISCUSSION Holger Hans Peter Freyther 2 participants 3 messages Jul 3, 2011

lore
Share & tools
lore
#0 lore raw

Hi all,

I managed to get my build directory into an inconsistent state. There was a kernel update and now the image fails to be built. It tries to copy the System.map-... from the sysroot/${MACHINE}/kernel/... but it is empty:

| FATAL: Could not open '/poky/build/tmp/sysroots/sysmocom-foo/kernel/System.map-2.6.33-07172-g57e3d92':
No such file or directory
| FATAL: Could not open '/poky/build/tmp/sysroots/sysmocom-foo/kernel/System.map-2.6.33-07172-g57e3d92':
No such file or directory
| FATAL: Could not open '/poky/build/tmp/sysroots/sysmocom-foo/kernel/System.map-2.6.33-07172-g57e3d92':
No such file or directory
| FATAL: Could not open '/poky/build/tmp/sysroots/sysmocom-foo/kernel/System.map-2.6.33-07172-g57e3d92':
No such file or directory

I tried a bitbake -cclean virtual/kernel, also a cleanall, interestingly the sysmocom-ti/kernel directory is not removed or does not seem to be re-populated. Did anyone else ever see that behavior? There was one non standard thing in the kernel upgrade:

+PR... bumped
-SRC_URI = "git://linux.git;protocol=ssh;branch=master" +SRC_URI = "git://linux.git;protocol=ssh;branch=master \
+ file://defconfig-${MACHINE}"
S = "${WORKDIR}/git"

do_configure() {

-       oe_runmake foo_defconfig
-       oe_runmake oldconfig
+       install -m 0644 ${WORKDIR}/defconfig-${MACHINE} ${S}/.config
+       #oe_runmake foo_defconfig
+       #oe_runmake oldconfig

}

is there anything that makes click?

holger

#1 lore raw
Quoted 39 lines · Holger Hans Peter Freyther

Hi all,

I managed to get my build directory into an inconsistent state. There was a kernel update and now the image fails to be built. It tries to copy the System.map-... from the sysroot/${MACHINE}/kernel/... but it is empty:

| FATAL: Could not open '/poky/build/tmp/sysroots/sysmocom-foo/kernel/System.map-2.6.33-07172-g57e3d92':
No such file or directory
| FATAL: Could not open '/poky/build/tmp/sysroots/sysmocom-foo/kernel/System.map-2.6.33-07172-g57e3d92':
No such file or directory
| FATAL: Could not open '/poky/build/tmp/sysroots/sysmocom-foo/kernel/System.map-2.6.33-07172-g57e3d92':
No such file or directory
| FATAL: Could not open '/poky/build/tmp/sysroots/sysmocom-foo/kernel/System.map-2.6.33-07172-g57e3d92':
No such file or directory

I tried a bitbake -cclean virtual/kernel, also a cleanall, interestingly the sysmocom-ti/kernel directory is not removed or does not seem to be re-populated. Did anyone else ever see that behavior? There was one non standard thing in the kernel upgrade:

(quoted diff) +6 -3
+PR... bumped
-SRC_URI = "git://linux.git;protocol=ssh;branch=master"
+SRC_URI = "git://linux.git;protocol=ssh;branch=master \
+          file://defconfig-${MACHINE}"
 S = "${WORKDIR}/git"
 
 do_configure() {
-       oe_runmake foo_defconfig
-       oe_runmake oldconfig
+       install -m 0644 ${WORKDIR}/defconfig-${MACHINE} ${S}/.config
+       #oe_runmake foo_defconfig
+       #oe_runmake oldconfig
 }
Quoted 1 line

is there anything that makes click?

I'm reminded of this commit:

http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=2b8a68001ada2885f15ca68aa7c94cc92bb36973

Did you try a "bitbake -ccleansstate virtual/kernel"?

I'm wondering if stale sstate files for the kernel are affecting the build somehow? Are you using a custom do_install/copy of kernel.bbclass or anything like that?

Cheers,

Richard

#2 lore raw
Quoted 5 lines · Richard Purdie

Did you try a "bitbake -ccleansstate virtual/kernel"?

I'm wondering if stale sstate files for the kernel are affecting the build somehow? Are you using a custom do_install/copy of kernel.bbclass or anything like that?

Hi Richard,

thanks a lot. It turns out I am stupid and I have had a shell mistake in my modified rootfs_rpm, it is still interesting who tries to access the System.map (I don't have the kernel-abi file, so image.bbclass should not try), and that FATAL actually is not fatal here.

Does anyone know if bash can return the line it exited from? I think would be great if our logs would include something like. BASH stopped processing at line... in case of errors.

sorry for the noise.

holger