Fails to build on i386: LLVM ERROR: out of memory #2600
Comments
nunotexbsd
added
bug
yasminetalby
added
investigating
|
Hello @nunotexbsd , Thank you very much for your submission. Unfortunately this is a compiler issue. Do you need the EC2Client? This might help resolve your issue. Let me know if you have more questions or request. Best regards, Yasmine |
yasminetalby
added
response-requested
|
Hello @yasminetalby , FreeBSD port is configured to build all clients by default.
Is there a quick way of listing available clients? |
github-actions
bot
removed
the
response-requested
|
Hello @nunotexbsd , Thank you very much for your response and for sharing more information on FreeBSD port configuration. You can list available clients using Let me know if you have any further questions! Best regards, Yasmine |
|
Hey @nunotexbsd we were discussing this, and yes it is related to the fact that we pushed out a change that seems push EC2 over the limit that requires more space to build. What memory constraints does your build machine have? This has been a reoccurring issue as the size of the EC2 code base grows, we are currently looking into how to break up the client to avoid this issue. In the meantime if you use a non-unity build your build should succeeded i.e. by specifying Let me know if that is a viable workaround for you while we work on finding ways to reduce the size of the EC2 footprint. |
|
Hello @sbiscigl , Same error with Thanks |
|
@nunotexbsd thanks for giving that a shot, we get complaints about the EC2 client a lot so we decided to break up the clients during unity build compile in hopes that the memory usage for compile will go down. see pull/2617. This will break larger clients into several smaller cpp files so that we place the burden of the larger client on the linker. So if you could please give todays release(to be release later today) going back to the original submission
did you see it work on other 32 bit archs? |
|
Same error with d5e2ed0 patch. I will test it with and without patch on armv7. |
|
136 builds OK on i386 arch. Thanks! |
sorry i forgot to mention one very important thing that i think you already found out in your testing and that is with the patch, the clients causing the problems weren't actually regenerated into smaller components. they were regenerated in the release process so the release tagged at 136 will actually have the fix, but not the patch commit. that only changed the process in which they were generated. But sounds like its working let me know if you see any of your archs failing, if not, then i think hopefully we can close this. we've been meaning to do this for a while, so thanks for giving us a little motivation to do so. |
|
@sbiscigl Logs: Please reopen pr if you need me to test more patches/fixes on armv7. Thanks very much |
|
Describe the bug
I'm updating from 114 to 130 and build fails on i386 (didn't check other 32 bit archs yet).
I did build all tags from 115 to 118 OK and it start failing on 119.
Logs:
118
119
130
Attached files:
c++: note: diagnostic msg: /tmp/ub1_EC2-4223dd.cpp
c++: note: diagnostic msg: /tmp/ub1_EC2-4223dd.sh
Expected Behavior
Build ok
Current Behavior
Fails on i386.
Reproduction Steps
build on i386 arch
Possible Solution
No response
Additional Information/Context
No response
AWS CPP SDK version used
1.11.119 / 1.11.130
Compiler and Version used
Clang 13.0.0
Operating System and version
FreeBSD 12.4 i386
The text was updated successfully, but these errors were encountered: