Have builds print the SHAs being used about openj9-openjdk-jdk HOT 10 CLOSED

I personally think it should be the other way (eg build time) - doing partial builds is par for the course doing JIT development and I worry that if the version is not computed at build time doubt about what is actually in a build will creep in which will cost development time and needless confusion - right now we can trust the version in the string is the version that was used to build 100% and I worry the configure-time option loses that fidelity and unambiguous trust we can place in the SHAs today.

from openj9-openjdk-jdk.

Do you think it should be the other way around (openjdk SHA is computed at build time)?

from openj9-openjdk-jdk.

Do you think it should be the other way around (openjdk SHA is computed at build time)?

+1 from me. Having all the SHAs reflect what was built makes the most sense. While both approaches are acceptable for the Adopt builds as they start from a clean slate each time, I agree with Andrew that incremental builds are more common for developer workflows. Better to accurately list the SHAs in -version to avoid confusion when comparing builds.

from openj9-openjdk-jdk.

@keithc-ca

from openj9-openjdk-jdk.

As I started working on this, I noticed that the openjdk SHA is computed only at configuration time, while the openj9 and omr SHAs are computed at build time. I plan to remove this inconsistency by computing all three SHAs (only) at configuration time. Printing the results at configuration time should then be sufficient.

from openj9-openjdk-jdk.

Here's a sample of the new configuration output (for jdk11):

checking for m4... /usr/bin/m4
Source versions:
  openjdk - 7e023948526
  openj9  - 015d3630dec
  omr     - e2fac34fcb3
checking CUDA_HOME... /home/keithc/opt/cuda-11.0

from openj9-openjdk-jdk.

I plan to remove this inconsistency by computing all three SHAs (only) at configuration time.

This may be confusing for developers who are updating the source and re-building.

from openj9-openjdk-jdk.

I'm not familiar with all dev workflows but I just wanted to say that build time sha is not necessarily 100% accurate. The files can still be modified by hand and not committed to Git. If you wanted 100% you could build into your SHA check a Git status check and print some kind of asterisk * with the SHA if the status was not clean. But maybe that's an edge case not worth solving?

from openj9-openjdk-jdk.

If we agree to use build-time SHAs (that seems to be the consensus we're approaching), then eclipse-openj9/openj9#9658 will want more attention so those incremental builds don't take longer than they should.

from openj9-openjdk-jdk.

This can be closed once these are merged:

• ibmruntimes/openj9-openjdk-jdk8#455
• ibmruntimes/openj9-openjdk-jdk11#357
• ibmruntimes/openj9-openjdk-jdk15#40
• #244

from openj9-openjdk-jdk.