I've added some comments inline to clarify.
Greg Von Kuster wrote:
Hi
Lance,
I need to figure out precisely what steps you are taking to
produce this behavior, as I have not been able to do so. Please see my
inline comments, and let me know more information about each step you
are taking to produce this behavior.
On Sep 25, 2012, at 10:24 AM, Lance Parsons wrote:
I'm sorry I wasn't more
clear. I do believe that those links explain the behavior I am seeing.
However, let me try to describe it a different way. It seems that there
will, at most, be one installable revision of a given version of a
tool. Here I use revision to denote a mercurial revision and version to
denote the version described in the tool's config xml. However, while
it seems this is the desired behavior, it seems to lead to an
undesirable situation in some circumstances. Consider this scenario:
1) Upload version 0.1 of MYTOOL at revision 0:XXXXXXXX to the tool shed.
I
assume you are uploading the initial tarball or separate files from a
certain local directory. Let's call it TOOL_DEV_LOCAL_DIR.
Not quite. I would typically create the repository using the Tool Shed
web interface. Then, on my local workstation, I would create the
TOOL_DEV_LOCAL_DIR by using the 'hg clone
http://toolshedurl/repo'
command.
2) Install version 0.1 of MYTOOL at revision 0:XXXXXXX to my local
Galaxy instance.
This now creates a
new mercurial repository at a specified location on your Galaxy server,
let's call it REPO_INSTALL_DIR.
Yes, I use the Galaxy Admin interface to install the tool from a tool
shed to my Galaxy instance.
3) Modify MYTOOL, leaving the version at 0.1,
You
are doing the above step in TOOL_DEV_LOCAL_DIR, correct?
Yes, I modify the files in TOOL_DEV_LOCAL_DIR, leaving the version the
same.
but
changing the revision
to 1:YYYYYYYY.
The tool
shed is creating the new change set hash for you when you push the
changes from TOOL_DEV_LOCAL_DIR to the repository in the tool shed,
correct?
I don't think so. I would typically commit the changes by issuing an
'hg commit' while in my TOOL_DEV_LOCAL_DIR and then push them to the
tool shed using the 'hg push' command.
These updates would be for things that don't change
the
input or output of a tool (such as updates to documentation, the
addition of tool dependencies, etc.)
OK,
this make sense.
Right, I think we're on the same page here, though this should probably
be spelled out somewhere in a "Best Practices" type document for tool
development (which could also help to clarify the steps to setup a sane
development environment, test tools, and push them to the public tool
shed. I've not even quite gotten a solid workflow for myself, so
publishing one might be useful.
4) Attempt to update MYTOOL from my local Galaxy instance.
You
are using the Galaxy UI to "Get updates" to your cloned repository,
which is located in REPO_INSTALL_DIR, correct?
Yes, then I use the Galaxy web admin interface to "Get updates".
This
now
results in an exception: " RepoLookupError: unknown revision '1' ".
If answers to all
of my above questions are "Yes", then this is where I cannot reproduce
the behavior you are seeing. If you are actually adding changsets to
your cloned repository located in REPO_INSTALL_DIR, then pushing them to
the repository in the tool shed, then attempting to get updates from
the tool shed to your cloned repository in REPO_INSTALL_DIR, then this
process may be causing the exception you are seeing. You should not be
adding changesets to the cloned repository, but to your local
development environment in TOOL_DEV_LOCAL_DIR.
I don't add changesets to the REPO_INSTALL_DIR, but instead make changes
in TOOL_DEV_LOCAL_DIR, commit them using mercurial, and then push them
to the tool shed using hg push.
Deactivating
and/or uninstalling MYTOOL works, but any attempt to
reinstall the latest version results in an error stating that an older
revision of the tool was installed. I now have to way to (easily)
update my local Galaxy instance to this updated revision of the tool.
My only workaround to this was to increment the version of MYTOOL and
push another revision. This allowed me to install the new version and
deactivate the old version in my local Galaxy instance.
As for the mercurial issue, I'm running OS X, using the Homebrew version
of Mercurial:
hg --version
Mercurial Distributed SCM (version 2.3.1)
(see
http://mercurial.selenic.com
for more information)
Copyright (C) 2005-2012 Matt Mackall and others
This is free software; see the source for copying conditions. There is
NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR
PURPOSE.
I'm running
version 2.2.3, so I'll upgrade to version 2.3.1 to see if the newer
version breaks my environment.
--
Lance Parsons - Scientific Programmer
134 Carl C. Icahn Laboratory
Lewis-Sigler Institute for Integrative Genomics
Princeton University