On 3/26/2020 3:18 PM, Denys Dmytriyenko wrote:
On Fri, Mar 06, 2020 at 09:53:46AM -0500, Jacob Stiffler wrote:
On 3/5/2020 7:15 PM, Denys Dmytriyenko wrote:
On Thu, Mar 05, 2020 at 04:49:05PM -0500, Jacob Stiffler wrote:
On 3/5/2020 4:24 PM, Denys Dmytriyenko wrote:
Jake,
Somehow this update has a really hard time getting built - it failed in the
same place with what looks like an our-of-memory error:
| src_generated/open62541/namespace0_generated.c: In function
‘namespace0_generated’:
| src_generated/open62541/namespace0_generated.c:113178:15: note: variable
tracking size limit exceeded with ‘-fvar-tracking-assignments’, retrying without
| 113178 | UA_StatusCode namespace0_generated(UA_Server *server) {
| | ^
| arm-none-linux-gnueabihf-gcc: fatal error: Killed signal terminated program
lto1
| compilation terminated.
| lto-wrapper: fatal error:
/opt/gcc-arm-9.2-2019.12-x86_64-arm-none-linux-gnueabihf/bin/arm-none-linux-gnueabihf-gcc
returned 1 exit status
| compilation terminated.
|
/opt/gcc-arm-9.2-2019.12-x86_64-arm-none-linux-gnueabihf/bin/../lib/gcc/arm-none-linux-gnueabihf/9.2.1/../../../../arm-none-linux-gnueabihf/bin/ld:
error: lto-wrapper failed
It was not specific to any platform or a build node - all platforms on all
build nodes have failed. I had to revert the update for now, as I needed to
release rc3 very quickly.
Has this been tested? Any comments?
I have built this for both am57xx-evm and am65xx-evm and nativesdk. I have
also run-time tested this on x86 and am65x IDK.
I did not ice that it took a significantly longer time to build with this
update and noticed "lto1-ltrans" as the top hog of resources. I have only
noticed this within OE. I do not notice any performance difference in the
standalone build.
I suspect this may be a result of enabling UA_NAMESPACE_ZERO=FULL, but I do
not know why this is isolated to OE.
Thanks. Is it possible to disable UA_NAMESPACE_ZERO=FULL to see if it improves
the build resource demands? Is it one of the required features to be enabled?
This is something we want because it is required to make use of other
standard model namespaces, such as OpcUaDi (device integration) and others.
I can move this to the PACKAGECONFIG so it is more easily configured, but
eventually we will need to determine how to make this work in our builds.
Jake,
Any more updates on this one?
Unfortunately, the old version now breaks on master with this error:
cc1: error: unrecognized command line option ‘-Wno-static-in-inline’ [-Werror]
And I started looking into this update again, but it's taking a very-very long
time even when lto is disabled...
I was waiting for a response or confirmation on the PACKAGECONFIG
approach. I have noticed some variation in the build times, and it seems
that it may be related to the toolchain. It seems that nativesdk builds
much faster than target, but I haven't experimented enough to be sure.
If it sounds OK, I can move the NAMESPACE_ZERO to a PACKAGECONFIG nad
leave it disabled by default.
_______________________________________________
meta-arago mailing list
[email protected]
http://arago-project.org/cgi-bin/mailman/listinfo/meta-arago
