Skip to content

Conversation

whitequark
Copy link
Member

No description provided.

This is preparing ground for a future commit where the testsuite
can link plugins from a non-installed ocamlbuild tree.
This results in spurious rebuilds and broken parallel builds.
Before this commit, it was very easy to accidentally use artifacts
from the installed ocamlbuild instance. In fact, it was impossible
to test linking of plugins properly, because plugins were always
linked with the system ocamlbuildlib.cm{x,}a.

After this commit, the testsuite does not need any user action
to use the correct ocamlbuild executable or ocamlbuildlib.cm{x,}a.
A specific testsuite can be invoked with `make test-NAME` (where NAME
is currently one of internal, findlibonly, external), and every
testsuite can be invoked with `make test`.

Also, any failing test now sets the exit status of the testsuite
to 1.
@whitequark
Copy link
Member Author

@gasche Any objections here? I've redone the OCAMLBUILDLIB change to be much less fragile, both internally and externally (see 0c24b7d).

@gasche
Copy link
Member

gasche commented Feb 27, 2017

No objection at all in principle. I had the impression that I saw a preliminary version of this PR with only the .travis.yml file, without the preliminary commits, and I wanted to convince myself that it actually worked without the other change. I may have just dreamed, or you may have rebased, but this looks good. (I have not tested the Makefile/testsuite changes, but I trust you that they work.)

@gasche
Copy link
Member

gasche commented Feb 27, 2017

(There is this mesmerizing "Some checks haven't completed yet" that makes me want to wait, but feel free to merge now, or I'll do it when I see the CI result.)

@gasche
Copy link
Member

gasche commented Feb 27, 2017

So the test fails, of course, given that mtime fails. @whitequark is it your intention to have a failing CI as long as your fixing PR is not merged or the change reverted, or should we temporarily disable the mtime part? I am fine with either.

@whitequark
Copy link
Member Author

I may have just dreamed, or you may have rebased

You've seen that. I rebase a lot, in general. Keeps the history clean.

is it your intention to have a failing CI as long as your fixing PR is not merged

That is my intention indeed, because master is broken right now. No reason to hide that status.

@whitequark whitequark merged commit 8a4b892 into ocaml:master Feb 27, 2017
@whitequark whitequark deleted the travis branch February 27, 2017 20:59
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants