after chocolatey 0.10.0 release automated deploy fails

[anuriq]

after upgrade to chocolatey 0.10.0 packages with no checksums will ask
for confirmation, which of course breaks automated deployment.
to work around the issue introduce an attribute switched on by default
to pass --allow-empty-checksums to choco executable. later when most
of packages will be updated we can switch off this by default.
more info here: chocolatey/choco#910

[mwrock]

I think the intent of the chocolatey change was to ensure package consumers are aware of potential threats and provide a means of "opting out" via the --allow-empty-checksums flag.

Having the chef cookbook subvert this seems like the wrong direction. You can certainly pass in any flag via the resource using options. If you want to globally enable this flag, you could use a batch or powershell script resource and execute choco feature enable -n allowEmptyChecksums.

[ferventcoder]

I've already made @anuriq aware, but for anyone else that happens in here, the confirmation prompt does not, I repeat, the prompt does NOT in any way break automated deployment.

Chocolatey has a custom PowerShell host that it runs (and it only asks for confirmation when that host is detected). That host is special in that Prompt for confirmation, Read-Host and Prompt for choice do not block the console and work in completely automated scenarios.

In the image, please note "Respond within 30 seconds" - also notice that no answer was provided and how it moved forward with the default selection.

This was implemented in 0.9.10.0. See chocolatey/choco#8 and also https://github.com/chocolatey/choco/blob/master/CHANGELOG.md#0910-june-17-2016

[ferventcoder]

@legal90 @Kasen ^^

[ferventcoder]

Now that said, packages that use HTTP without checksums fail to install in 0.10.0. But there are options there.

[anuriq]

@mwrock @ferventcoder my intent was to make a quick fix, so that there would be no need to edit existing cookbooks relying on chocolatey. then, when most of the choco packages are updated to include checksum, this fix would be removed. but if this really doesn't fit, then decline the PR. i guess everyone is able to find a workaround.