Skip to contentSkip to conversation

gcc10: Don't default back to -fcommon

PATCH In review Jacob Kroon 5 participants 25 messages Aug 4, 2020 – Aug 18, 2020

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 gcc10: Don't default back to -fcommon

It has been almost 3 months since the upgrade to gcc 10. Switch back to relying on gcc default configuration (-fno-common).

This reverts OE-Core commit 951e859b1e8297970278c539e989b8a6d06a9cb3.

---
 meta/recipes-devtools/gcc/gcc-configure-common.inc | 1 -
 1 file changed, 1 deletion(-)
meta/recipes-devtools/gcc/gcc-configure-common.inc +0 -1
index 6cb40fcb74..a64c4caf00 100644
@@ -34,7 +34,6 @@ EXTRA_OECONF = "\
     --program-prefix=${TARGET_PREFIX} \
     --without-local-prefix \
     --disable-install-libiberty \
-    --with-specs=%{!fno-common:%{!fcommon:-fcommon}} \
     ${EXTRA_OECONF_BASE} \
     ${EXTRA_OECONF_GCC_FLOAT} \
     ${EXTRA_OECONF_PATHS} \
#1 lore raw
Quoted 4 lines · Jacob Kroon

It has been almost 3 months since the upgrade to gcc 10. Switch back to relying on gcc default configuration (-fno-common).

have you done a world build with this? In core+meta-oe I still see 70+ recipes failing
what will be good is if people start sending fixes before switching the defaults.

Quoted 20 lines

This reverts OE-Core commit 951e859b1e8297970278c539e989b8a6d06a9cb3.

---
 meta/recipes-devtools/gcc/gcc-configure-common.inc | 1 -
 1 file changed, 1 deletion(-)
meta/recipes-devtools/gcc/gcc-configure-common.inc +0 -1
index 6cb40fcb74..a64c4caf00 100644
@@ -34,7 +34,6 @@ EXTRA_OECONF = "\
     --program-prefix=${TARGET_PREFIX} \
     --without-local-prefix \
     --disable-install-libiberty \
-    --with-specs=%{!fno-common:%{!fcommon:-fcommon}} \
     ${EXTRA_OECONF_BASE} \
     ${EXTRA_OECONF_GCC_FLOAT} \
     ${EXTRA_OECONF_PATHS} \
#2 lore raw
Quoted 11 lines · Khem Raj
Quoted 4 lines · Jacob Kroon

It has been almost 3 months since the upgrade to gcc 10. Switch back to relying on gcc default configuration (-fno-common).

have you done a world build with this? In core+meta-oe I still see 70+ recipes failing
what will be good is if people start sending fixes before switching the defaults.

Ah, that I haven't, only the images I use. Is there a list of failing recipes available somewhere ?

Jacob

#3 lore raw
Quoted 8 lines · Jacob Kroon
Quoted 1 line · Khem Raj

On Tue, Aug 4, 2020 at 1:20 AM Jacob Kroon <[email protected]>

Quoted 5 lines · wrote:
Quoted 4 lines

It has been almost 3 months since the upgrade to gcc 10. Switch back to relying on gcc default configuration (-fno-common).

Quoted 8 lines
Quoted 5 lines

have you done a world build with this? In core+meta-oe I still see 70+ recipes failing
what will be good is if people start sending fixes before switching the defaults.

Ah, that I haven't, only the images I use. Is there a list of failing recipes available somewhere ?

I think I posted the list when I submitted the gcc10 upgrade but this patch is already in master-next so Richard will soon share the Oe-core list atleast

Quoted 4 lines

Jacob

#4 lore raw

here is a recent build with clang11 which also switched to using -fno-common and errors are pretty much same you will see with gcc10 with this patch to revert the -fcommon defaults.

https://errors.yoctoproject.org/Errors/Build/107234/?limit=100

I was waiting for upstream packages to get fixed before switching the defaults back to -fno-common.

Quoted 27 lines · Khem Raj
Quoted 8 lines · Jacob Kroon
Quoted 6 lines · Khem Raj
Quoted 4 lines · Jacob Kroon

It has been almost 3 months since the upgrade to gcc 10. Switch back to relying on gcc default configuration (-fno-common).

Quoted 8 lines
Quoted 5 lines

have you done a world build with this? In core+meta-oe I still see 70+ recipes failing
what will be good is if people start sending fixes before switching the defaults.

Ah, that I haven't, only the images I use. Is there a list of failing recipes available somewhere ?

I think I posted the list when I submitted the gcc10 upgrade but this patch is already in master-next so Richard will soon share the Oe-core list atleast

Quoted 4 lines

Jacob

#5 lore raw
Quoted 9 lines · Khem Raj

here is a recent build with clang11 which also switched to using -fno-common and errors are pretty much same you will see with gcc10 with this patch to revert the -fcommon defaults.

https://errors.yoctoproject.org/Errors/Build/107234/?limit=100

I was waiting for upstream packages to get fixed before switching the defaults back to -fno-common.

Yes I understand, maybe its better to hold this for now.

Having a list of recipes that needs fixing would probably speed things up. I can have a look at some of them.

Jacob

#6 lore raw
Quoted 12 lines · Jacob Kroon
Quoted 10 lines · Khem Raj

here is a recent build with clang11 which also switched to using -fno-common and errors are pretty much same you will see with gcc10 with this patch to revert the -fcommon defaults.

https://errors.yoctoproject.org/Errors/Build/107234/?limit=100

I was waiting for upstream packages to get fixed before switching the
defaults back to -fno-common.

Quoted 4 lines

Yes I understand, maybe its better to hold this for now.

Having a list of recipes that needs fixing would probably speed things up. I can have a look at some of them.

I ran this through with just OE-Core and the list of failures is:

https://autobuilder.yoctoproject.org/typhoon/#/builders/52/builds/2237

I'm a little frustrated that patches like this get sent and them people like Khem and I have to spot and avoid the problems as there isn't testing :/.

I'll drop the patch for now until more of these issues are resolved.

Cheers,

Richard

#7 lore raw
In reply to Khem Raj
Quoted 8 lines · Khem Raj

here is a recent build with clang11 which also switched to using -fno-common and errors are pretty much same you will see with gcc10 with this patch to revert the -fcommon defaults.

https://errors.yoctoproject.org/Errors/Build/107234/?limit=100

I was waiting for upstream packages to get fixed before switching the defaults back to -fno-common.

Khem requested I revisit this. I see at least:

a) perf failures pretty much everywhere
b) kexec-tools in no-x11 build
c) cpio build failures in testimage

https://autobuilder.yoctoproject.org/typhoon/#/builders/83/builds/1264

the world builds haven't completed yet.

So looks much better but still a few things need fixes.

Cheers,

Richard

#8 lore raw
Quoted 6 lines · Richard Purdie
Quoted 4 lines · Khem Raj

here is a recent build with clang11 which also switched to using -fno-common and errors are pretty much same you will see with gcc10 with this patch to revert the -fcommon defaults.

Quoted 8 lines
Quoted 4 lines

https://errors.yoctoproject.org/Errors/Build/107234/?limit=100

I was waiting for upstream packages to get fixed before switching the defaults back to -fno-common.

Khem requested I revisit this. I see at least:

a) perf failures pretty much everywhere

This is interesting

| x86_64-poky-linux-gcc: error: unrecognized command-line option '-m64/include/uapi/asm-generic/errno.h'

I wonder if its some kernel update which is causing this, it seems to be missing a space in commandline somewhere

Quoted 1 line

b) kexec-tools in no-x11 build

I think I dealt with this in meta-oe perhaps we need to use -fcommon in cflags.

Quoted 2 lines

c) cpio build failures in testimage

cpio needs a patch to build see
https://git.openembedded.org/openembedded-core/commit/?id=ca6e5511b59b585de793ff6424c7c6ccfcd8e61e
I guess we need to either patch cpio with this patch on target before building or pass -fcommon to the test so this patch is not needed

Thanks for the run, I will try to look into it, I think for cpio build failure perhaps we need to pass

Quoted 11 lines

https://autobuilder.yoctoproject.org/typhoon/#/builders/83/builds/1264

the world builds haven't completed yet.

So looks much better but still a few things need fixes.

Cheers,

Richard

#9 lore raw

Bruce,

We need https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a
backported to 5.4 kernels for perf to build with -fno-common can you
queue this in your kernel updates as soon as you can ?

Thanks
-Khem

Quoted 51 lines · Khem Raj
Quoted 6 lines · Richard Purdie
Quoted 4 lines · Khem Raj

here is a recent build with clang11 which also switched to using -fno-common and errors are pretty much same you will see with gcc10 with this patch to revert the -fcommon defaults.

Quoted 8 lines
Quoted 4 lines

https://errors.yoctoproject.org/Errors/Build/107234/?limit=100

I was waiting for upstream packages to get fixed before switching the defaults back to -fno-common.

Khem requested I revisit this. I see at least:

a) perf failures pretty much everywhere

This is interesting

| x86_64-poky-linux-gcc: error: unrecognized command-line option '-m64/include/uapi/asm-generic/errno.h'

I wonder if its some kernel update which is causing this, it seems to be missing a space in commandline somewhere

Quoted 1 line

b) kexec-tools in no-x11 build

I think I dealt with this in meta-oe perhaps we need to use -fcommon in cflags.

Quoted 2 lines

c) cpio build failures in testimage

cpio needs a patch to build see
https://git.openembedded.org/openembedded-core/commit/?id=ca6e5511b59b585de793ff6424c7c6ccfcd8e61e
I guess we need to either patch cpio with this patch on target before building or pass -fcommon to the test so this patch is not needed

Thanks for the run, I will try to look into it, I think for cpio build failure perhaps we need to pass

Quoted 11 lines

https://autobuilder.yoctoproject.org/typhoon/#/builders/83/builds/1264

the world builds haven't completed yet.

So looks much better but still a few things need fixes.

Cheers,

Richard

#11 lore raw
Quoted 13 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 2 lines

backported to 5.4 kernels for perf to build with -fno-common can you queue this in your kernel updates as soon as you can ?

I think we have other active kernels which will also need it as even with Khem's patch we still see failures:

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

Cheers,

Bruce

Quoted 7 lines

https://autobuilder.yoctoproject.org/typhoon/#/builders/83/builds/1267

Cheers,

Richard

Signature · 3 lines
- Thou shalt not follow the NULL pointer, for chaos and madness await thee
at its end
- "Use the force Harry" - Gandalf, Star Trek II
#12 lore raw

Awesome, thanks!

Quoted 37 lines · Bruce Ashfield
Quoted 4 lines · Richard Purdie
Quoted 2 lines · Khem Raj

Bruce,

Quoted 8 lines
Quoted 4 lines

We need
https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a
backported to 5.4 kernels for perf to build with -fno-common can you queue this in your kernel updates as soon as you can ?

I think we have other active kernels which will also need it as even with Khem's patch we still see failures:

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

Cheers,

Bruce

Quoted 7 lines

https://autobuilder.yoctoproject.org/typhoon/#/builders/83/builds/1267

Cheers,

Richard

--
- Thou shalt not follow the NULL pointer, for chaos and madness await thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

#13 lore raw
In reply to Bruce Ashfield
Quoted 14 lines · Bruce Ashfield
Quoted 10 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 1 line

backported to 5.4 kernels for perf to build with -fno-common can

you

Quoted 1 line

queue this in your kernel updates as soon as you can ?

Quoted 7 lines
Quoted 4 lines

I think we have other active kernels which will also need it as even
with Khem's patch we still see failures:

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

I bumped the meta-yocto BSPs with a quickly hacked up patch but we still see:

https://autobuilder.yoctoproject.org/typhoon/#/builders/65/builds/2339

so I think there may be another patch we need somewhere...

(which matches what we saw with Khem's _append which also failed).

There is still some piece of this we're missing, maybe on non-IA.

Cheers,

Richard

#14 lore raw
Quoted 21 lines · Richard Purdie
Quoted 10 lines · Bruce Ashfield
Quoted 6 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 9 lines
Quoted 8 lines
Quoted 1 line

backported to 5.4 kernels for perf to build with -fno-common can

you

Quoted 1 line

queue this in your kernel updates as soon as you can ?

I think we have other active kernels which will also need it as even
with Khem's patch we still see failures:

Quoted 7 lines
Quoted 2 lines

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

I bumped the meta-yocto BSPs with a quickly hacked up patch but we still see:

I have local bumps for the yocto-bsps as well, let me try a perf build against beaglebone to see if I can see anything.

If I build master-next, will I get all the required gcc patches to trigger the issue ?

Bruce

Quoted 13 lines

https://autobuilder.yoctoproject.org/typhoon/#/builders/65/builds/2339

so I think there may be another patch we need somewhere...

(which matches what we saw with Khem's _append which also failed).

There is still some piece of this we're missing, maybe on non-IA.

Cheers,

Richard

Signature · 3 lines
- Thou shalt not follow the NULL pointer, for chaos and madness await thee
at its end
- "Use the force Harry" - Gandalf, Star Trek II
#15 lore raw
Quoted 33 lines · Bruce Ashfield
Quoted 22 lines · Richard Purdie
Quoted 10 lines · Bruce Ashfield
Quoted 6 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 1 line
Quoted 1 line
Quoted 1 line

backported to 5.4 kernels for perf to build with -fno-common

can

Quoted 8 lines
Quoted 7 lines

you

Quoted 1 line

queue this in your kernel updates as soon as you can ?

I think we have other active kernels which will also need it as even
with Khem's patch we still see failures:

Quoted 6 lines
Quoted 2 lines

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

I bumped the meta-yocto BSPs with a quickly hacked up patch but we still see:

Quoted 5 lines

I have local bumps for the yocto-bsps as well, let me try a perf build against beaglebone to see if I can see anything.

If I build master-next, will I get all the required gcc patches to trigger the issue ?

You need:

http://git.yoctoproject.org/cgit.cgi/poky/commit/?h=master-next&id=c79101ffff8a9cc70daefe737e16b773190bb6ca

without its revert (which is also in -next atm).

Cheers,

Richard

#16 lore raw
In reply to Richard Purdie
Quoted 24 lines · Richard Purdie
Quoted 14 lines · Bruce Ashfield
Quoted 10 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 1 line

backported to 5.4 kernels for perf to build with -fno-common can

you

Quoted 1 line

queue this in your kernel updates as soon as you can ?

Quoted 7 lines
Quoted 4 lines

I think we have other active kernels which will also need it as even
with Khem's patch we still see failures:

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

Quoted 9 lines

I bumped the meta-yocto BSPs with a quickly hacked up patch but we still see:

https://autobuilder.yoctoproject.org/typhoon/#/builders/65/builds/2339

so I think there may be another patch we need somewhere...

(which matches what we saw with Khem's _append which also failed).

there seems to be concaternation of compiler options without space

| arm-poky-linux-gnueabi-gcc: error: unrecognized argument in option '-mfpu=neon/include/uapi/asm-generic/errno.h'
| arm-poky-linux-gnueabi-gcc: note: valid arguments to '-mfpu=' are:
auto crypto-neon-fp-armv8 fp-armv8 fpv4-sp-d16 fpv5-d16 fpv5-sp-d16 neon neon-fp-armv8 neon-fp16 neon-vfpv3 neon-vfpv4 vfp vfp3 vfpv2 vfpv3 vfpv3-d16 vfpv3-d16-fp16 vfpv3-fp16 vfpv3xd vfpv3xd-fp16 vfpv4 vfpv4-d16
| arm-poky-linux-gnueabi-gcc: fatal error: no input files

but it also showed up on x86_64 in one of your prior builds, so I wonder if its something in kernel build system or is it processing TUNE_CCARGS in special ways.

Quoted 6 lines

There is still some piece of this we're missing, maybe on non-IA.

Cheers,

Richard

#17 lore raw
Quoted 48 lines · Khem Raj
Quoted 24 lines · Richard Purdie
Quoted 14 lines · Bruce Ashfield
Quoted 10 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 1 line

backported to 5.4 kernels for perf to build with -fno-common can

you

Quoted 1 line

queue this in your kernel updates as soon as you can ?

Quoted 7 lines
Quoted 4 lines

I think we have other active kernels which will also need it as even
with Khem's patch we still see failures:

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

Quoted 9 lines

I bumped the meta-yocto BSPs with a quickly hacked up patch but we still see:

https://autobuilder.yoctoproject.org/typhoon/#/builders/65/builds/2339

so I think there may be another patch we need somewhere...

(which matches what we saw with Khem's _append which also failed).

there seems to be concaternation of compiler options without space

| arm-poky-linux-gnueabi-gcc: error: unrecognized argument in option '-mfpu=neon/include/uapi/asm-generic/errno.h'
| arm-poky-linux-gnueabi-gcc: note: valid arguments to '-mfpu=' are:
auto crypto-neon-fp-armv8 fp-armv8 fpv4-sp-d16 fpv5-d16 fpv5-sp-d16 neon neon-fp-armv8 neon-fp16 neon-vfpv3 neon-vfpv4 vfp vfp3 vfpv2 vfpv3 vfpv3-d16 vfpv3-d16-fp16 vfpv3-fp16 vfpv3xd vfpv3xd-fp16 vfpv4 vfpv4-d16
| arm-poky-linux-gnueabi-gcc: fatal error: no input files

Quoted 6 lines

but it also showed up on x86_64 in one of your prior builds, so I wonder if its something in kernel build system or is it processing TUNE_CCARGS in special ways.

Quoted 2 lines

There is still some piece of this we're missing, maybe on non-IA.

Perhaps these kernel backports might help

https://github.com/torvalds/linux/commit/7597ce89b3ed239f7a3408b930d2a6c7a4c938a1
https://github.com/torvalds/linux/commit/6d1f916265aa615218d341ba5d3fd39dfe931cde

Quoted 4 lines
Quoted 4 lines

Cheers,

Richard

#18 lore raw
Quoted 57 lines · Khem Raj
Quoted 7 lines · Khem Raj
Quoted 4 lines · Richard Purdie
Quoted 2 lines · Bruce Ashfield
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 6 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 33 lines
Quoted 12 lines
Quoted 11 lines
Quoted 8 lines
Quoted 1 line

backported to 5.4 kernels for perf to build with -fno-common can

you

Quoted 1 line

queue this in your kernel updates as soon as you can ?

I think we have other active kernels which will also need it as even
with Khem's patch we still see failures:

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

Quoted 9 lines

I bumped the meta-yocto BSPs with a quickly hacked up patch but we still see:

https://autobuilder.yoctoproject.org/typhoon/#/builders/65/builds/2339

so I think there may be another patch we need somewhere...

(which matches what we saw with Khem's _append which also failed).

there seems to be concaternation of compiler options without space

| arm-poky-linux-gnueabi-gcc: error: unrecognized argument in option '-mfpu=neon/include/uapi/asm-generic/errno.h'
| arm-poky-linux-gnueabi-gcc: note: valid arguments to '-mfpu=' are:
auto crypto-neon-fp-armv8 fp-armv8 fpv4-sp-d16 fpv5-d16 fpv5-sp-d16 neon neon-fp-armv8 neon-fp16 neon-vfpv3 neon-vfpv4 vfp vfp3 vfpv2 vfpv3 vfpv3-d16 vfpv3-d16-fp16 vfpv3-fp16 vfpv3xd vfpv3xd-fp16 vfpv4 vfpv4-d16
| arm-poky-linux-gnueabi-gcc: fatal error: no input files

Quoted 6 lines

but it also showed up on x86_64 in one of your prior builds, so I wonder if its something in kernel build system or is it processing TUNE_CCARGS in special ways.

Quoted 2 lines

There is still some piece of this we're missing, maybe on non-IA.

Quoted 6 lines

Perhaps these kernel backports might help

https://github.com/torvalds/linux/commit/7597ce89b3ed239f7a3408b930d2a6c7a4c938a1

https://github.com/torvalds/linux/commit/6d1f916265aa615218d341ba5d3fd39dfe931cde

I unfortunately had to clean and rebuild my development machine, so only now started a reference h/w build to see what's up with perf.

I'll follow up tomorrow when I have some input.

Cheers,

Bruce

Quoted 7 lines
Quoted 6 lines
Quoted 6 lines

Cheers,

Richard

Signature · 3 lines
- Thou shalt not follow the NULL pointer, for chaos and madness await thee
at its end
- "Use the force Harry" - Gandalf, Star Trek II
#19 lore raw
In reply to Khem Raj

Yes, I'm seeing this weird concatenation with raspberrypi builds as well:

find: unknown predicate `-mthumb/arch' arm-webos-linux-gnueabi-gcc: error: unrecognized command-line option '-mthumb/include/uapi/asm-generic/errno.h' arm-webos-linux-gnueabi-gcc: fatal error: no input files compilation terminated.

But in this case it isn't fatal for the perf build, I've noticed it only because it was failing with 5.4.51 currently in meta-raspberrypi and after upgrade to 5.4.58 perf built OK, but this error message is still there in log.do_compile.

Quoted 60 lines · Khem Raj
Quoted 4 lines · Richard Purdie
Quoted 2 lines · Bruce Ashfield
Quoted 8 lines
Quoted 8 lines
Quoted 6 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 12 lines
Quoted 11 lines
Quoted 8 lines
Quoted 1 line

backported to 5.4 kernels for perf to build with -fno-common can

you

Quoted 1 line

queue this in your kernel updates as soon as you can ?

I think we have other active kernels which will also need it as even
with Khem's patch we still see failures:

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

Quoted 9 lines

I bumped the meta-yocto BSPs with a quickly hacked up patch but we still see:

https://autobuilder.yoctoproject.org/typhoon/#/builders/65/builds/2339

so I think there may be another patch we need somewhere...

(which matches what we saw with Khem's _append which also failed).

there seems to be concaternation of compiler options without space

| arm-poky-linux-gnueabi-gcc: error: unrecognized argument in option '-mfpu=neon/include/uapi/asm-generic/errno.h'
| arm-poky-linux-gnueabi-gcc: note: valid arguments to '-mfpu=' are:
auto crypto-neon-fp-armv8 fp-armv8 fpv4-sp-d16 fpv5-d16 fpv5-sp-d16 neon neon-fp-armv8 neon-fp16 neon-vfpv3 neon-vfpv4 vfp vfp3 vfpv2 vfpv3 vfpv3-d16 vfpv3-d16-fp16 vfpv3-fp16 vfpv3xd vfpv3xd-fp16 vfpv4 vfpv4-d16
| arm-poky-linux-gnueabi-gcc: fatal error: no input files

but it also showed up on x86_64 in one of your prior builds, so I wonder if its something in kernel build system or is it processing TUNE_CCARGS in special ways.

Quoted 6 lines

There is still some piece of this we're missing, maybe on non-IA.

Cheers,

Richard

#20 lore raw
Quoted 8 lines · Martin Jansa

Yes, I'm seeing this weird concatenation with raspberrypi builds as well:

find: unknown predicate `-mthumb/arch' arm-webos-linux-gnueabi-gcc: error: unrecognized command-line option '-mthumb/include/uapi/asm-generic/errno.h' arm-webos-linux-gnueabi-gcc: fatal error: no input files compilation terminated.

confirmed as the same with my beaglebone build that ran overnight.

I'll have a closer look and send a fix (hopefully) .. but if anyone else is already doing this, please let me know and I'll work on something else!

Cheers,

Bruce

Quoted 69 lines

But in this case it isn't fatal for the perf build, I've noticed it only because it was failing with 5.4.51 currently in meta-raspberrypi and after upgrade to 5.4.58 perf built OK, but this error message is still there in log.do_compile.

Quoted 60 lines · Khem Raj
Quoted 4 lines · Richard Purdie
Quoted 2 lines · Bruce Ashfield
Quoted 8 lines
Quoted 8 lines
Quoted 6 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 12 lines
Quoted 11 lines
Quoted 8 lines
Quoted 1 line

backported to 5.4 kernels for perf to build with -fno-common can

you

Quoted 1 line

queue this in your kernel updates as soon as you can ?

I think we have other active kernels which will also need it as even
with Khem's patch we still see failures:

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

Quoted 9 lines

I bumped the meta-yocto BSPs with a quickly hacked up patch but we still see:

https://autobuilder.yoctoproject.org/typhoon/#/builders/65/builds/2339

so I think there may be another patch we need somewhere...

(which matches what we saw with Khem's _append which also failed).

there seems to be concaternation of compiler options without space

| arm-poky-linux-gnueabi-gcc: error: unrecognized argument in option '-mfpu=neon/include/uapi/asm-generic/errno.h'
| arm-poky-linux-gnueabi-gcc: note: valid arguments to '-mfpu=' are:
auto crypto-neon-fp-armv8 fp-armv8 fpv4-sp-d16 fpv5-d16 fpv5-sp-d16 neon neon-fp-armv8 neon-fp16 neon-vfpv3 neon-vfpv4 vfp vfp3 vfpv2 vfpv3 vfpv3-d16 vfpv3-d16-fp16 vfpv3-fp16 vfpv3xd vfpv3xd-fp16 vfpv4 vfpv4-d16
| arm-poky-linux-gnueabi-gcc: fatal error: no input files

but it also showed up on x86_64 in one of your prior builds, so I wonder if its something in kernel build system or is it processing TUNE_CCARGS in special ways.

Quoted 6 lines

There is still some piece of this we're missing, maybe on non-IA.

Cheers,

Richard

Signature · 3 lines
- Thou shalt not follow the NULL pointer, for chaos and madness await thee
at its end
- "Use the force Harry" - Gandalf, Star Trek II
#21 lore raw

The perf issue with find and cflags concatenation is fixed in upstream by:
https://github.com/torvalds/linux/commit/e4ffd066ff440a57097e9140fa9e16ceef905de8

Adding V=1 to perf's EXTRA_OEMAKE makes it easier to see, sending a patch now.

Quoted 101 lines · Bruce Ashfield
Quoted 8 lines · Martin Jansa

Yes, I'm seeing this weird concatenation with raspberrypi builds as well:

find: unknown predicate `-mthumb/arch' arm-webos-linux-gnueabi-gcc: error: unrecognized command-line option '-mthumb/include/uapi/asm-generic/errno.h' arm-webos-linux-gnueabi-gcc: fatal error: no input files compilation terminated.

confirmed as the same with my beaglebone build that ran overnight.

I'll have a closer look and send a fix (hopefully) .. but if anyone else is already doing this, please let me know and I'll work on something else!

Cheers,

Bruce

Quoted 69 lines

But in this case it isn't fatal for the perf build, I've noticed it only because it was failing with 5.4.51 currently in meta-raspberrypi and after upgrade to 5.4.58 perf built OK, but this error message is still there in log.do_compile.

Quoted 60 lines · Khem Raj
Quoted 4 lines · Richard Purdie
Quoted 2 lines · Bruce Ashfield
Quoted 8 lines
Quoted 8 lines
Quoted 6 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 12 lines
Quoted 11 lines
Quoted 8 lines
Quoted 1 line

backported to 5.4 kernels for perf to build with -fno-common can

you

Quoted 1 line

queue this in your kernel updates as soon as you can ?

I think we have other active kernels which will also need it as even
with Khem's patch we still see failures:

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

Quoted 9 lines

I bumped the meta-yocto BSPs with a quickly hacked up patch but we still see:

https://autobuilder.yoctoproject.org/typhoon/#/builders/65/builds/2339

so I think there may be another patch we need somewhere...

(which matches what we saw with Khem's _append which also failed).

there seems to be concaternation of compiler options without space

| arm-poky-linux-gnueabi-gcc: error: unrecognized argument in option '-mfpu=neon/include/uapi/asm-generic/errno.h'
| arm-poky-linux-gnueabi-gcc: note: valid arguments to '-mfpu=' are:
auto crypto-neon-fp-armv8 fp-armv8 fpv4-sp-d16 fpv5-d16 fpv5-sp-d16 neon neon-fp-armv8 neon-fp16 neon-vfpv3 neon-vfpv4 vfp vfp3 vfpv2 vfpv3 vfpv3-d16 vfpv3-d16-fp16 vfpv3-fp16 vfpv3xd vfpv3xd-fp16 vfpv4 vfpv4-d16
| arm-poky-linux-gnueabi-gcc: fatal error: no input files

but it also showed up on x86_64 in one of your prior builds, so I wonder if its something in kernel build system or is it processing TUNE_CCARGS in special ways.

Quoted 6 lines

There is still some piece of this we're missing, maybe on non-IA.

Cheers,

Richard

--
- Thou shalt not follow the NULL pointer, for chaos and madness await thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

#22 lore raw
Quoted 4 lines · Martin Jansa

The perf issue with find and cflags concatenation is fixed in upstream by:
https://github.com/torvalds/linux/commit/e4ffd066ff440a57097e9140fa9e16ceef905de8

this should fix it. thanks for nailing it, I guess older kernels will have tough time with gcc10

Quoted 96 lines

Adding V=1 to perf's EXTRA_OEMAKE makes it easier to see, sending a patch now.

Quoted 93 lines · Bruce Ashfield
Quoted 7 lines · Martin Jansa

Yes, I'm seeing this weird concatenation with raspberrypi builds as well:

find: unknown predicate `-mthumb/arch'
arm-webos-linux-gnueabi-gcc: error: unrecognized command-line option '-mthumb/include/uapi/asm-generic/errno.h' arm-webos-linux-gnueabi-gcc: fatal error: no input files
compilation terminated.

confirmed as the same with my beaglebone build that ran overnight.

I'll have a closer look and send a fix (hopefully) .. but if anyone else is already doing this, please let me know and I'll work on something else!

Cheers,

Bruce

Quoted 64 lines

But in this case it isn't fatal for the perf build, I've noticed it only because it was failing with 5.4.51 currently in meta-raspberrypi and after upgrade to 5.4.58 perf built OK, but this error message is still there in log.do_compile.

Quoted 59 lines · Khem Raj
Quoted 24 lines · Richard Purdie
Quoted 14 lines · Bruce Ashfield
Quoted 10 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 1 line

backported to 5.4 kernels for perf to build with -fno-common can

you

Quoted 1 line

queue this in your kernel updates as soon as you can ?

Quoted 7 lines
Quoted 4 lines

I think we have other active kernels which will also need it as even
with Khem's patch we still see failures:

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

Quoted 9 lines

I bumped the meta-yocto BSPs with a quickly hacked up patch but we still see:

https://autobuilder.yoctoproject.org/typhoon/#/builders/65/builds/2339

so I think there may be another patch we need somewhere...

(which matches what we saw with Khem's _append which also failed).

there seems to be concaternation of compiler options without space

| arm-poky-linux-gnueabi-gcc: error: unrecognized argument in option '-mfpu=neon/include/uapi/asm-generic/errno.h'
| arm-poky-linux-gnueabi-gcc: note: valid arguments to '-mfpu=' are:
auto crypto-neon-fp-armv8 fp-armv8 fpv4-sp-d16 fpv5-d16 fpv5-sp-d16 neon neon-fp-armv8 neon-fp16 neon-vfpv3 neon-vfpv4 vfp vfp3 vfpv2 vfpv3 vfpv3-d16 vfpv3-d16-fp16 vfpv3-fp16 vfpv3xd vfpv3xd-fp16 vfpv4 vfpv4-d16
| arm-poky-linux-gnueabi-gcc: fatal error: no input files

but it also showed up on x86_64 in one of your prior builds, so I wonder if its something in kernel build system or is it processing TUNE_CCARGS in special ways.

Quoted 6 lines

There is still some piece of this we're missing, maybe on non-IA.

Cheers,

Richard

--
- Thou shalt not follow the NULL pointer, for chaos and madness await thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

#23 lore raw
In reply to Martin Jansa
Quoted 8 lines · Martin Jansa

The perf issue with find and cflags concatenation is fixed in upstream by:

https://github.com/torvalds/linux/commit/e4ffd066ff440a57097e9140fa9e16ceef905de8

Adding V=1 to perf's EXTRA_OEMAKE makes it easier to see, sending a patch now.

Yah. I did the same thing here, was chasing the issue through the generated files, but hadn't tracked down the root cause yet.

Looks good to me. I'll chuck away my debug.

Bruce

Quoted 105 lines
Quoted 102 lines · Bruce Ashfield
Quoted 8 lines · Martin Jansa

Yes, I'm seeing this weird concatenation with raspberrypi builds as well:

find: unknown predicate `-mthumb/arch' arm-webos-linux-gnueabi-gcc: error: unrecognized command-line option '-mthumb/include/uapi/asm-generic/errno.h' arm-webos-linux-gnueabi-gcc: fatal error: no input files compilation terminated.

confirmed as the same with my beaglebone build that ran overnight.

I'll have a closer look and send a fix (hopefully) .. but if anyone else is already doing this, please let me know and I'll work on something else!

Cheers,

Bruce

Quoted 70 lines

But in this case it isn't fatal for the perf build, I've noticed it only because it was failing with 5.4.51 currently in meta-raspberrypi and after upgrade to 5.4.58 perf built OK, but this error message is still there in log.do_compile.

Quoted 61 lines · Khem Raj
Quoted 4 lines · Richard Purdie
Quoted 2 lines · Bruce Ashfield
Quoted 8 lines
Quoted 8 lines
Quoted 6 lines · Richard Purdie
Quoted 4 lines · Khem Raj

Bruce,

We need

https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a

Quoted 9 lines
Quoted 9 lines
Quoted 8 lines
Quoted 1 line

backported to 5.4 kernels for perf to build with -fno-common can

you

Quoted 1 line

queue this in your kernel updates as soon as you can ?

I think we have other active kernels which will also need it as even
with Khem's patch we still see failures:

Quoted 7 lines
Quoted 2 lines

I've taken care of linux-yocto now, I don't think there are any active use cases for those kernels that won't be covered.

I bumped the meta-yocto BSPs with a quickly hacked up patch but we still see:

https://autobuilder.yoctoproject.org/typhoon/#/builders/65/builds/2339

Quoted 5 lines

so I think there may be another patch we need somewhere...

(which matches what we saw with Khem's _append which also failed).

there seems to be concaternation of compiler options without space

| arm-poky-linux-gnueabi-gcc: error: unrecognized argument in option '-mfpu=neon/include/uapi/asm-generic/errno.h'
| arm-poky-linux-gnueabi-gcc: note: valid arguments to '-mfpu=' are:
auto crypto-neon-fp-armv8 fp-armv8 fpv4-sp-d16 fpv5-d16 fpv5-sp-d16 neon neon-fp-armv8 neon-fp16 neon-vfpv3 neon-vfpv4 vfp vfp3 vfpv2 vfpv3 vfpv3-d16 vfpv3-d16-fp16 vfpv3-fp16 vfpv3xd vfpv3xd-fp16 vfpv4 vfpv4-d16
| arm-poky-linux-gnueabi-gcc: fatal error: no input files

but it also showed up on x86_64 in one of your prior builds, so I wonder if its something in kernel build system or is it processing TUNE_CCARGS in special ways.

Quoted 6 lines

There is still some piece of this we're missing, maybe on non-IA.

Cheers,

Richard

--
- Thou shalt not follow the NULL pointer, for chaos and madness await thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

Signature · 3 lines
- Thou shalt not follow the NULL pointer, for chaos and madness await thee
at its end
- "Use the force Harry" - Gandalf, Star Trek II
#24 lore raw
In reply to Richard Purdie
Quoted 9 lines · Richard Purdie
Quoted 6 lines · Khem Raj

Bruce,

We need
https://github.com/torvalds/linux/commit/168200b6d6ea0cb5765943ec5da5b8149701f36a
backported to 5.4 kernels for perf to build with -fno-common can you queue this in your kernel updates as soon as you can ?

Quoted 5 lines

I think we have other active kernels which will also need it as even with Khem's patch we still see failures:

https://autobuilder.yoctoproject.org/typhoon/#/builders/83/builds/1267

I see

meta_ide.MetaIDE.test_meta_ide_can_build_cpio_project: FAILED (64.00s)

It seems the sdk fix for cpio is not working properly

Quoted 4 lines

Cheers,

Richard