packaging: dynamically define make targets
We currently have make boilerplate for each instance of a distro's
release. This is redundant, annoying to maintain, and prone to errors.
This commit defines variables holding available releases for
various distros. We then iterate through the list and dynamically
define make targets.
Differential Revision: https://phab.mercurial-scm.org/D3761
packaging: don't write files for templatized Dockerfiles
Now that Docker image building is implemented in Python and we
can perform template substitution in memory, we don't need to
write out produced Dockerfiles to disk.
Differential Revision: https://phab.mercurial-scm.org/D3760
packaging: replace dockerlib.sh with a Python script
I want to do some more advanced things with Docker in upcoming
commits. Trying to do that with shell scripts will be a bit too
painful for my liking. Implementing things in Python will be
vastly simpler in the long run.
This commit essentially ports dockerlib.sh to a Python script.
dockerdeb and dockerrpm have been ported to use the new hg-docker
script.
hg-docker requires Python 3. I've only tested on Python 3.5.
Unlike the local packaging scripts which may need to run on old
distros, the Docker packaging scripts don't have these constraints.
So I think it is acceptable to require Python 3.5.
As part of the transition, the Docker image tags changed slightly.
I don't think that's a big deal: the Docker image names are
effectively arbitrary and are a means to an end to achieve
running commands in Docker containers.
The code for resolving the Dockerfile content allows substituting
values passed as arguments. This will be used in a subsequent commit.
Differential Revision: https://phab.mercurial-scm.org/D3759
packaging: consistently create build user in Dockerfiles
Previously, dockerlib.sh appended some commands to create a
"build" user in each Docker image. The resulting Docker images
could be inconsistent depending on the execution environment
and base image.
With this change, we explicitly create our custom user and
group as the first action in each Dockerfile. The user always
has user:group 1000:1000 and all built images are consistent.
We also create a home directory for the user under /build.
This directory is currently ignored.
As part of this, we stop setting the DBUILDUSER variable in
dockerlib.sh and instead set it in the respective scripts that
call it. This is in preparation for further refactoring of
dockerlib.sh.
Differential Revision: https://phab.mercurial-scm.org/D3758
scmutil: move construction of instability count message to separate fn
When the commad we are running, introduces new instabilities, we show a message
like `5 new orphan changesets`, `2 new content-divergent changesets`, `1 new
phase-divergent changesets` etc which is very nice.
Now taking a step ahead, we want users to show how to fix them too. Something
like:
`5 new orphan changesets (run 'hg evolve' to resolve/stabilize them)`
`2 new content-divergent changesets (run 'hg evolve --content-divergent' to
resolve them)`
and maybe telling user a way to understand more about those new instabilities
like `hg evolve --list` or `hg log -r 'orphan()'` something like that.
The idea came from
issue5855 which I want to fix because fixing that will result
in a nice UI.
Taking the construction logic out will allow extensions like evolve (maybe
rebase too) to wrap that and add information about how to resolve and how to
understand the instability more.
Differential Revision: https://phab.mercurial-scm.org/D3734
procutil: use unbuffered stdout on Windows
Windows doesn't support line buffering, treating it as fully buffered. This
causes output of slow commands to stutter. We use unbuffered instead.