- Sep 12, 2019
-
-
Georges Racinet authored
-
Georges Racinet authored
It'll be nicer for simulation mode summary
-
Georges Racinet authored
To build and push everything, not running the container in the end would look like ./build-launch --build-base-image --no-run --push We aren't so far from a CI goal that would launch the tests and push only if they do pass.
-
Georges Racinet authored
Helpful if we just want to build
-
Georges Racinet authored
One less hardcoded constant can only help in the long run.
-
Georges Racinet authored
what I remembered incorrectly as Docker environment is actually Docker context, as any build command will tell you in its output.
-
Georges Racinet authored
Just running is already helpful; also, in conjunction with --image-tag, this provides an quick way to check the base image.
-
Georges Racinet authored
The option was there, but the corresponding 'if' statement had simply been forgotten.
-
Georges Racinet authored
Removing duplication is always a good thing, and in particular, makes these sequences of operations controlled by boolean flags easier to follow. Here this exposes a bug.
-
- Sep 11, 2019
-
-
Georges Racinet authored
There are more merge requests to bless like this for Heptapod 0.6.0, this is just an intermediate step.
-
Georges Racinet authored
The old launch.sh wasn't so useful without volume mounts, and it was obviously out of the question to introduce paths that make sense on a given developer system only. Rewriting in Python makes it much more auto-documented and easily configurable, thanks to the power of argparse, and also allows to have defaults that could be acceptable for many developers. Some of the defaults (e.g., HTTP port) are tailored for heptapod-tests defaults. Finally, all of this makes sense for developer installations, so the new script is within heptapod_docker_dev/
-
Georges Racinet authored
I don't want to depend on it, because docker-compose is not universal. Besides, for it to be useful in daily development or deployment, it would need to include installation specific stuff that would be at best confusing for general users.
-
Georges Racinet authored
It is a pain for people that need to test their changes easily, and dev.heptapod.net has no problem handling such small pulls, especially with the `pull_branch` restriction.
-
- Aug 01, 2019
-
-
Georges Racinet authored
I'm mostly tired of setting these manually on each update from my mounted host repos.
-
- Sep 08, 2019
-
-
Georges Racinet authored
- the explanation for 'unreleased' is more global - subsections for better user awareness of breaking changes
-
Georges Racinet authored
-
- Sep 07, 2019
-
-
Georges Racinet authored
The changelog for the unreleased version contains only those changes that are available in the current latest Docker image tag
-
Georges Racinet authored
Docker Hub's markdown doesn't like it
-
Georges Racinet authored
-
- Sep 06, 2019
-
-
Georges Racinet authored
This is not yet a formal merge of the 10.1.7+ce.0 tag in omnibus, but this doesn't change much, the only actual change is a bump of Mattermost to 4.2.2
-
Georges Racinet authored
See CHANGES.md for details
-
- Apr 28, 2019
-
-
Georges Racinet authored
It's been forgotten in e768d8cd04c7, which then wasn't so trimmed
-
- Sep 06, 2019
-
-
Georges Racinet authored
All component updates were actually still using the path stored during the initial cloning Did not notice this earlier because I usually wait for the mirror to be up to date.
-
- Sep 02, 2019
-
-
Georges Racinet authored
For the time being, it's simpler to act on both sides of that pipe in one repo, and besides, this is actually where we used to copy inside the image.
-
- Aug 30, 2019
-
-
Georges Racinet authored
We'd like to test this a bit in practice before releasing it.
-
Georges Racinet authored
This is done after the fact from Mercurial logs. From now on the goal will be to fill this before tagging new releases.
-
- Aug 28, 2019
-
-
Georges Racinet authored
This leftover development shortcut had the effect that the write permission was always transmitted to the hg backend. This is less of a security hole than it looks to be, because most permission checking is done before hand in the Rails controller (as heptapod-tests@842f6acda290 demonstrate)
-
- Aug 13, 2019
-
-
Georges Racinet authored
This is a step before a possible 0.5.2 Issue heptapod#81 (aka hg-git#3) is worse than I thought, because obsmarkers exchange is always total, making it happen if the prunes seem to be unrelated to what's been pushed.
- Aug 12, 2019
-
-
Georges Racinet authored
We just add the fix in hg-git for heptapod#65. All other components are as in 0.5.0
-
- Aug 05, 2019
-
-
Georges Racinet authored
For next release, we'll give this hg-git changeset a tag. In the meanwhile, this will allow us to test in the wild with the octobus/heptapod:latest image.
-
- Aug 02, 2019
-
- Aug 01, 2019
-
-
Georges Racinet authored
This is for development images.
-
Georges Racinet authored
This is doubly ironic, because the whole purpose of this file is to be persistent, and because the use of a variable is meant to avoid typos in the filename.
-
Georges Racinet authored
In `hg_path()` it is perfectly normal that the required path alias doesn't exist. Without this change, this used to output "Not found!" to the console, easily confused with a problem in update to revision.
-
- Jul 19, 2019
-
- Jul 09, 2019
-
- Jul 08, 2019
-
-
Georges Racinet authored
With some additional configuration, this will give us proper logs and the capability to send them to Sentry.
-
Georges Racinet authored
By default, in version 19.9.0, Gunicorn master process would capture standard outputs of the workers and won't reoutput them to its own. We're sending them to a separate file anyway.
-