Get Rid of nobuild and nolink#2765
Conversation
|
Actually, I think that |
|
@adamjstewart Do you know why Travis tests have stalled here (or how to restart)? Thanks! |
|
Not sure why they stalled. But yes, pushing a new commit is one way to get them to restart. So is closing and reopening the PR. I think if you go to the travis page linked to under details, there is sometimes a restart button as well. |
|
Hmm... pushing the new commit didn't work. Can't find a restart button. now trying closing/reopening. |
|
Still no tests :*( |
|
This passed all the tests; except Travis hung and hasn't yet started the Mac tests. I'm adding the "ready" label... |
|
It's not hung -- it's just waiting for the mac worker. I suspect those are scarce, but I'd rather have mac tests than not. |
|
The mac tests take quite a while. If you check in |
|
Ok, so it takes a while to find an available mac worker, and the tests take a while to run |
|
I understand it may take a while... I'm still betting optimistically that
the test will succeed.
…On Fri, Jan 6, 2017 at 2:04 PM, Todd Gamblin ***@***.***> wrote:
It's not hung -- it's just waiting for the mac worker. I suspect those are
scarce, but I'd rather have mac tests than not.
—
You are receiving this because you modified the open/close state.
Reply to this email directly, view it on GitHub
<#2765 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AB1cd_6mKCAZXhpItuZVZQedrZbWU-2Rks5rPpBSgaJpZM4Lc5-H>
.
|
|
P.S. This may conflict a lot with #2761. |
|
They modify different lines; so I think both PRs will go in with few or no
conflicts.
But if there are conflicts... we'll figure it out then.
…On Fri, Jan 6, 2017 at 2:28 PM, Adam J. Stewart ***@***.***> wrote:
P.S. This may conflict a lot with #2761
<#2761>.
—
You are receiving this because you modified the open/close state.
Reply to this email directly, view it on GitHub
<#2765 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AB1cd_VJZLLVeMMz3FnwIg3Iiik8cmszks5rPpXzgaJpZM4Lc5-H>
.
|
lib/spack/docs/packaging_guide.rst
Outdated
| @@ -1508,10 +1508,7 @@ Additional hybrid dependency types are (note the lack of quotes): | |||
| * **<not specified>**: ``type`` assumed to be ``("build", | |||
| "link")``. This is the common case for compiled language usage. | |||
| * **alldeps**: All dependency types. **Note:** No quotes here | |||
There was a problem hiding this comment.
alldeps can probably be eliminated from the user docs and packages, as well. It's used internally by the Spec traversal routines, which need a way to guarantee that they are going over the whole DAG.
Using alldeps in a package is going be very surprising when we add things like test dependencies. There are 4 packages that currently use alldeps -- those should probably be changed to ('build', 'link', 'run').
citibeth
force-pushed
the
efischer/170105-EliminateNolink
branch
from
7734683 to
7977854
Compare
citibeth
force-pushed
the
efischer/170105-EliminateNolink
branch
from
7977854 to
5882cdc
Compare
|
Closing; the SHA Travis don't match the SHA on the source branch ( |
|
@citibeth: I think you are seeing the effect of Travis merging the PR into the target branch before building. If you go to the Travis PR build history and click on one of the SHAs you see there, it takes you to a page like this one. You see there that the travis commit is ec1c44f, which is a merge of 5882cdc into the tip of |
|
@tgamblin OK, looks like it's OK now. My trouble was actually rather prosaic: I had neglected to commit the most recent change from my working tree. So it was not in Git, and Travis could not see it. And no amount of pushing and re-pushing my branch could change that. |
|
@tgamblin I have addressed your change request. |
3 participants
The dependency type aliases
nobuildandnolinkhave been deprecated. This PR gets rid of them.@tgamblin @hegner Would you be able to please take a quick look at the chnages to
database.py? I think they need a little help...