Feature - Rejected SVN Dependencies

[fronobulax]

There is some discussion in https://kolmafia.us/threads/crimbo-script.27060/ that suggests there are script failures because a dependency is present, but not the latest version. I looked at the code and AFAIK a dependency is checked only for presence. If a version of the dependency is already installed then there is no attempt to update the dependency.

Users with svnUpdateOnLogin=true will probably never see this. We could change the dependency action from install to install or update but that might be a problem for people with svnUpdateOnLogin=false because they want more control over updates.

We can document the behavior or we can change it. I am wary of even trivial changes to the SVN code, for several reasons, but if there is interest in autoupdating dependencies regardless of svnUpdateOnLogin then that can probably be done.

 

[Magus_Prime]

I think that there is more than one thing at work here.

As noted in the thread: I ran svn update prior to installing the script in question. While, according to the text in the gCLI, an update to autoscend was installed at that time it seems that "something" happened to make that not quite true. This is the issue to which I alluded in https://kolmafia.us/threads/mafia-has-broken-something-post-r25995.26976/#post-165991

The above has nothing to do with the issue of "dependency is checked only for presence. If a version of the dependency is already installed then there is no attempt to update the dependency."

 

[fronobulax]

Noted.

This conversation is occurring in several places

My goal is to identify a situation where the failure to update is a problem that can be fixed by changes to KoLmafia or documentation.

So far the following are the SVN failures I know about.

There is a failure to update caused by communications with the remote repository. Generally this is beyond mafia's control and the only solution is to retry knowing that this failure may be the root cause of a different failure.

There is a failure to update because the local mafia maintained repository is corrupt. Deleting and reinstalling (or using a non-mafia tool to "fix" the repository) is the usual fix. There may be an opportunity for mafia's messages, in this case, to be more verbose.

There is a failure to update because the owner of the remote repository has structured it in a way that mafia does not understand, allow, or was expecting. The general solution is to fix the remote repository and try again. For example, mafia's handling of file names that are not unique across subdirectories was changed to deal with cases where mafia needed to accommodate the remote repository's structure.

So I am trying to understand whether the reported issues are something new or whether they are covered by the above. I note that autoascend is involved and for better or worse the maintainers have done things that have pushed the envelope of mafia's capabilities. Is this another instance?

I note my personal frustration with reports that "it doesn't work" coupled with a lack of information from others or lack of understanding on my part.

 

[Magus_Prime]

I note my personal frustration with reports that "it doesn't work" coupled with a lack of information from others or lack of understanding on my part.

Completely understandable. (no pun intended) I lack the knowledge to expand on why "it doesn't work" reliably but Taltamir just posted in the other thread with technical information.

 

[fronobulax]

See https://kolmafia.us/threads/crimbo-script.27060/#post-166225

As a mafia dev I need to remember that there is a known problem with GitHub's SVN bridge and that autoascend devs have made some choices that highlight this problem. This is likely to persist until GitHub fixes things, autoascend restructures or KoLmafia adds native git support.

But it does seem as if there is nothing to do, now, other than me personally remembering this case when people express concerns about mafia SVN's reliability.