The sysinternals zip file was updated on the origin server (SHA256 of https://download.sysinternals.com/files/SysinternalsSuite.zip is currently 9C55CEF14E4CFD65CDD2CAA27F3A718D68BDB7B9BDA14A54139B08B40B96015D). Looking inside, a few binaries were modified on 2017-06-08. However, Microsoft did not update the "last modified" text on the download page (https://technet.microsoft.com/en-us/sysinternals/bb842062), so AU did not notice the changes and the Chocolatey package did not get updated.
Assuming this is a one-time oversight on Microsoft side and the au_GetLatest logic does not need to change, is there an easy way to force AU to regenerate the checksums and publish the package with an explicitly specified version number?
probably won't happen that often.
Why use a explicit version?
Why not just do a normal fix update? [AU sysinternals]?
Since we only use 3 parts of the version, that should be fine
As @AdmiringWorm noted, recheck is possible with [AU sysinternals] which would generate fix version. To force AU to update package on checksum change see for example how I handle cpu-z package:
https://github.com/majkinetor/au-packages/blob/master/cpu-z.install/update.ps1#L30-L37
Why use a explicit version?
Why not just do a normal fix update?
The versioning scheme for the package is "yyyy.mm.dd", so 2017.05.22.20170609 would look silly :-)
It could also imply a packaging fix only (no change in binaries), which is not the case here. That is why I think it would be best to release it as version 2017.06.08 (the latest modification date of the zipped binaries).
@majkinetor Neat trick. In this package we could go even further and compute the version based on binary modification dates.
2017.05.22.20170609 would look silly
Package fix notations looks silly no matter what the version scheme IMO, so I feel that doesn't really matter.
It could also imply a packaging fix only (no change in binaries),
Not really, it implies that something have changed, but no new version is available. Which is appropriate in this case.
That is why I think it would be best to release it as version 2017.06.08 (the latest modification date of the zipped binaries).
I'm against this as that would imply that a new version is actually available, which isn't reflected on the project page.
The project page does not show the current version number, it only claims "Updated: May 22, 2017" (which may as well refer to the project page itself). Whatever, I don't really care about it.
Meanwhile, going back to my original question and clarifying it: does the commit message syntax ([AU ...]) support forcing a specific version number? (It seems that AU itself supports it via $au_Version.)
it only claims "Updated: May 22, 2017"
Which is what we use as a version.
Meanwhile, going back to my original question and clarifying it: does the commit message syntax ([AU ...]) support forcing a specific version number? (It seems that AU itself supports it via $au_Version.)
yes it does, using [AU packageName:packageVersion]
Thanks, good to know.
As it seems I'm outvoted here, I'll just trigger AU to generate a fix version.
Something's weird.
cup -y sysinternals on my choco 0.10.5 it still tries to install 2017.05.22clist sysinternals shows 2017.5.22.20170609choco outdated shows 2017.05.22 as latestIs there a known issue with "05" sometimes being treated as newer than "5"?
That certainly is odd. There isn't any issues I'm aware of that treats 05 as newer than 5.
Just tried to install it inside a test environment (running choco v 0.10.6.1), and it grabbed the latest version for me.
Also tried installing the previous version (ignoring the checksum) then upgrade it, still works fine for me.
Upgraded my choco to 0.10.7 - no change.
Fresh install on a test VM with choco 0.10.3 - it installed 2017.05.22.
By the way, choco info sysinternals returns for me "0 packages found", both on the main system and on the VM. Does it work for you?
yup, just tried choco info sysinternals which worked fine for me.
But that is a known issue though. If I remember correctly, it's possibly related to a caching issue.
Alright, it came to its senses. Must have been a CDN/cache issue.
Most helpful comment
Alright, it came to its senses. Must have been a CDN/cache issue.