A new variable is exposed to templates, which holds the filename of the
template source. The primary use case is to be able to include a warning
message within the template. For example:
# Do not edit. This file auto-generated from {{ yadm.source }}.
The new test system is written with py.test. These tests are more
comprehensive, run faster by an order of magnitude, and are far more
maintainable. The tests themselves conform to PEP8.
`eval` has always been used to process the globs in `.yadm/encrypt`.
This is being removed, as there is a risk of executing "dirty" data
found in `.yadm/encrypt`.
Caveats of this change:
* Brace and tilde expansion will no longer work in `.yadm/encrypt`
* Paths with spaces must not be quoted anymore
Directories are created prior to merge during clone, and prior to any
Git command run.
This directly addresses CVE-2017-11353.
When cloning a repo which includes data in a .ssh or .gnupg directory,
if those directories do not exist at the time of cloning, yadm will
create the directories with mask 0700 prior to merging the fetched data
into the work-tree.
When running a Git command and .ssh or .gnupg directories do not exist,
create those directories with mask 0700 prior to running the Git
command. However, do not create those directories if
yadm.auto-private-dirs is false.
`uname -s` was already being executed every run for *cygwin* detection.
I've consolidated all of the OS detection into a single function. This
also fixed the problem of running `uname -s` twice for the `alt`
command.
With the new functionality, when the 'alt' command is called (or automatically
triggered), any file with a name ending in '##yadm_tmpl' is treated as a jinja
template. The template is processed by envtpl and the result is written to a
file without the '##yadm_tmpl' name. The variables passed into the template
processing are
YADM_CLASS
YADM_OS
YADM_HOSTNAME
YADM_USER
These variables are set according to the normal rules for
CLASS, OS, HOSTNAME, and USER during the alt processing.
