[oe] [meta-java][PATCH] openjdk-8: update aarch64 version to 8u162b12

André Draszik git at andred.net
Wed May 23 11:54:49 UTC 2018


On Wed, 2018-05-23 at 12:37 +0100, André Draszik wrote:
> It looks like the files have changed, as the checksums for this 162b12
> update aren't right anymore...
> 
> Starting to look into this, I wonder if the URLs used by the OpenJDK-8
> recipes are actually meant to be stable? The URL pattern is
> 
> http://hg.openjdk.java.net/${OPENJDK_ARCH_PORT}/${OPENJDK_HG_U}/archive/${
> OPENJDK_FILE}
> 
> Considering 'archive' in the URL, and considering you can download an
> archive of tip if you want to, and the fact that the server doesn't send
> the
> size in bytes of the file - are these created on demand, and can thusly
> change due to timestamps? (Similar to the github archive problems we're
> having).
> 
> On the other hand, the previous checksums for the 161b15 release are still
> valid.
> 
> I'd suggest to switch the OpenJDK-8 recipes to fetching from hg://
> directly,
> rather than using those archive URLs?
> 
> So, are those archives meant to be returning the same file forever? Any
> thoughts?

Luckily, I still had my originally downloaded file lying somewhere :-) So,
the only difference seems to be a file called '.hg_archival.txt' (all
timestamps for all files / directories within both new and old archives are
the same, even for the .hg_archival.txt file)

When I downloaded originally, it contained:

repo: cfeea66a3fa8ca3686a7cfa2d0ce8ab0169f168d
node: 917454da25c1735c001ad84fae01c61e0dac0f07
branch: default
latesttag: aarch64-jdk8u161-b15
latesttagdistance: 1

Whereas when I download now, it contains:

repo: cfeea66a3fa8ca3686a7cfa2d0ce8ab0169f168d
node: f8e58f4c29aecbadcd7f1ee0e28a4ab8c8a7e3b1
branch: default
tag: aarch64-jdk8u171-b10


Any thoughts why a snapshot of an older tag (162b12) would suddenly start to
reference a newer tag (171b10). And why does the snapshot of the even older
tag 161b15 still doesn't reference the 171b10 tag, even when downloaded
anew? I don't know enough about mercurial to make sense out of that...


Cheers,
Andre'




More information about the Openembedded-devel mailing list