Chocolatey-coreteampackages: (QuickTime) Find new location to check for updates.

[AdmiringWorm]

Considering how the quick time update process almost fails on every appveyor build run, I think we need to find another location to get the version and url from.

Anyone have any idea...

[majkinetor]

One of the things that could be done to minimize spam is:

try { 
  update
} catch { 
   on forbiden stuff just return nothing.
}

Alternative location would be better. Maybe we could utilize proxy for specific packages.

[AdmiringWorm]

the problem with that try catch is if there is any other errors, it won't be catched though.
But I think that is perhaps a lesser evil right now.

[AdmiringWorm]

I've added that try catch for now, but let's keep this open until an alternative location have been found.

[majkinetor]

You can catch specific errors and rethrow others.

[majkinetor]

try {1/0} catch { if ($_ -notmatch 'divide by zero') {throw $_} }
try {/0} catch { if ($_ -notmatch 'divide by zero') {throw $_} }

[majkinetor]

You should probably add some Write-Host in the script when you do this, so we can see that in report.

[AdmiringWorm]

so in this case it would probably be something like I think:

try { update } catch { if ($_ -notmatch "Access Denied") { throw $_ } }

[AdmiringWorm]

You should probably add some Write-Host in the script when you do this, so we can see that in report.

Sure, I'll add these later, busy with something else right now

[majkinetor]

try { update } catch { 
    if ($_ -notmatch "Access Denied") { throw $_ } }
    Write-Host 'Access denied error while checking version ignored, package might have an update'
}

It should be seen in output of updateall report, althoguh I didn't check it.

[AdmiringWorm]

I wonder if we should perhaps add something similiar for packages using sourceforge, as some of them fail quite often

[majkinetor]

Not a bad idea, IMO, although we should for start keep the number of such packages low to see how that approach behaves.

[AdmiringWorm]

👍
I'll add a try catch just to a couple of packages using sourceforge for now, and we can see how it works out.

[majkinetor]

But it should definitelly show up in a report for that package output so we can search for it.

[AdmiringWorm]

I've added it to the cdrtfe and jubler packages, I don't think we will see any difference until later tonight though.

[majkinetor]

I implemented changes in AU like this:

• update.ps1 can now return single word 'ignore'. It will not work if script returns anything else.
• you can provide output in the form of Write-Host and it will be seen in report

For example:

try {
    update -ChecksumFor 32
} catch {
    if ($_ -match 'meh 123') {   
         Write-Host  meh 123; 
         'ignore'   
    } else {throw $_}
}

should give

Path          : C:\work\au-packages\eac
Name          : eac
Updated       : False
Pushed        : False
RemoteVersion :
NuspecVersion : 1.3
Result        : {ignored

                eac - checking updates using au version 2016.12.1.192001
                meh 123
                }
Error         :
NuspecPath    : C:\work\au-packages\eac\eac.nuspec
NuspecXml     : #document

[majkinetor]

I updated quicktime package to test this. However the AV run fails entire build:

The running command stopped because the preference variable "ErrorActionPreference" or common parameter is set to Stop: 'Quicktime' update script didn't return AUPackage but: char

I honestly do not understand this behavior nor it happens on my system whatever I do, its AV specific. So I will have to test this for longer on my repository that has few packages outdating every now and then so it will take time.

[majkinetor]

Shell we close this ? As far as I can see the package passes every day.

[majkinetor]

Plus it now has ignore.

[AdmiringWorm]

yeah, we can close it now, although it would still be optimal to have an alternative location.