Hi all,
so as I saw you’re at one of the steps in the framework repo, I thought: “How
on earth did they get past 007?”
So I dug up the email, set my java version to the same as on the server and ran
the scripts …
ant -f releasesteps.xml Release_Step_007 -Drelease.version=0.9.7
-DskipTests=true
and the step failed with exactly the same error message as our attempts did:
notlast:
compare_files:
[echo] comparing royale-typedefs-node-0.9.7-typedefs.swc
BUILD FAILED
/Users/christofer.dutz/Projects/Apache/Royale/royale/royale-typedefs/releasesteps.xml:167:
The following error occurred while executing this line:
/Users/christofer.dutz/Projects/Apache/Royale/royale/royale-typedefs/releasesteps.xml:185:
The following error occurred while executing this line:
/Users/christofer.dutz/Projects/Apache/Royale/royale/royale-typedefs/releasesteps.xml:196:
The following error occurred while executing this line:
/Users/christofer.dutz/Projects/Apache/Royale/royale/royale-typedefs/releasesteps.xml:232:
royale-typedefs-node-0.9.7-typedefs.swc does not match
Total time: 1 minute 21 seconds
Even the differences were identical to mine (except that the timecode
differences seem to have been fixed or have gone away till the next time our
daylight-saving time is out of sync again)
So I thought: “Ok so this just used the release version of the compiler that I
created, so let’s update that to the one you folks staged”.
So I ran:
ant -f releasesteps.xml Release_Step_003 -Drelease.version=0.9.7
and
ant -f releasesteps.xml Release_Step_007 -Drelease.version=0.9.7
-DskipTests=true
after the first, so I was using the same compiler version. But it still failed.
When comparing the swcs of the node typedef IntelliJ tells me that the
catalog.xml of the node module built on the CI server contains a “http” script
(class) and the one built locally contains a “net” instead.
However the problem with the changed order of properties in the library.swc
seems to have been fixed by latest changes in the compiler.
So I thought … ok … let me check the others (the ant script aborts after the
first error).
And the “js” typedef has more differences:
* The SHA-256 checksums differ
* Document.cookie is String when built locally but was Object when built on
the CI server
* The number after the $ sign for quite some subclasses differs.
I gave up after that.
So I can’t really see this is a reproducible build.
However I don’t think they have to be, but if you insist on reproducible build,
I think they should be reproducible for anyone or they aren’t worth being
called “reproducible”.
Chris
