Forum OpenACS Q&A: 4.6 stalls on "Completing Install sequence"

Hi,

I recently tried to install 4.6 final.  It halts on "Completing Install sequence".  Log files shows an error of

ERROR:  apm_packages_package_key_fk referential integrity violation - key referenced from apm_packages not found in apm_package_types

I am posting here to help anyone that may encounter the problem in the future.  The basic problem is that acs-workflow is not included anymore in acs-core in CVS.  On the up and coming 4.7  it was decided that acs-workflow be not part of the core anymore.  So when I exported acs-core CVS module even with the revision of oacs-4-6-final it did not grab acs-workflow package.  This made acs-install.sql fail when its trying to install the packages which still included acs-workflow.

The workaround was simple enough to just add acs-workflow package.

I hope this helps someone in the future.  I do have a question, is the forums still the only place to things like this?

Collapse
Posted by Don Baccus on
Sure, posting here is fine.  But ... please don't expect the CVS HEAD to work in a mix-and-match way with the OpenACS 4.6 release ...
Collapse
Posted by Jeff Davis on
Don, It's not mix and match, the problem is a checkout of -r oacs-4-6-final acs-core
is not installable since we removed workflow from the
acs-core alias in CVS.  Hearing this,  I realize that it's
an issue and maybe we should put it back. We missed it since I
think we all do co opeancs-4 rather than acs-core when installing (certainly thats what I have been doing).
Collapse
Posted by Bart Teeuwisse on
Please, do put acs-workflow back into acs-core. To test dotLRN 1.0 we'll be checking out acs-core rather than openacs-4.

/Bart

Collapse
Posted by Jeff Davis on
I put acs-workflow back in the acs-core alias.
Collapse
Posted by Jun Yamog on
Maybe we can get help from some CVS guru.  How we can go about this problem.  Its a minor problem but maybe someone has a good solution for it.
Collapse
Posted by Jeff Davis on
Just check out acs-workflow in the packages directory if you don't already have it.
cd packages
cvs co acs-workflow
Collapse
Posted by Lars Pind on
This was my bad, sorry. Thanks, Jun, for locating the problem, and to Jeff for fixing it.

Jun, you may need to use "cvs up -d acs-workflow" to get the new directory?

/Lars

Collapse
Posted by Jun Yamog on
Hi,

Thanks for the advise, although I have to make myself clearer.  I was already able to fix the problem prior to posting the first message.  What I meant about asking a CVS guru on my last post, if there is better fix than our current fix.

As of now we have acs-workflow back into acs-core.  But logically when a user checkouts acs-core say on 4.7 or beyond he/she should not get acs-workflow anymore.  Its a minor thing, but maybe some people out there have an alternative solution.

A real drastic solution but illogical is to really create a repository for acs-core.  Packages will also have their own repository.  Right now we have a single tree for the core and packages right?

Anyway the current situation is fine and good, but if some knows a better solution please post.  Thanks.

Collapse
Posted by Don Baccus on
Bart ... dotLRN 1.0 should be tested against the OpenACS 4.6 release, not CVS HEAD.  That's the combination we'll be telling people to use (though we may end up pushing out the 4.6.1 bug-fix release if necessary).

dotLRN 1.1 will be tested against OpenACS 4.7 and will include the internationalization stuff.

Collapse
Posted by Lars Pind on
Why don't we simply change 4.6 so that no longer installs acs-workflow by default, either?

We already ascertained that it should be safe to do so, since all the packages that use acs-workflow already have it in their requires info.

Any complaints if I do this?

/Lars

Collapse
Posted by Jeff Davis on
Lars, yeah, go ahead and do it.  It was just expedient to
put it back in the acs-core alias until we were able to do that.
Collapse
Posted by Bart Teeuwisse on
Don,

dotLRN 1.0 will be tested against OpenACS 4.6. The CVS alias acs-core applies to all CVS branches including the OpenACS 4.6 branch.

I intend to use the OpenACS 4.6 branch to checkout a copy of OpenACS 4.6 (with bug fixes) when (re)installing dotLRN on the test servers.

We might find some bugs in OpenACS while testing dotLRN 1.0 and it seems logical to use the 4.6 branch to fix them.

/Bart