Age | Commit message (Collapse) | Author |
|
|
|
|
|
|
|
On Linux, 'attach' uses named pipes, and 'remote_console' uses -remsh.
The latter is usually deemed better since named pipes require a call to
fsync on every line written.
On Windows, no named pipes are available so attach uses -remsh directly.
Historically, remote_console was added to linux *after* attach, but no
alias was added for it on windows. Since there's a predominance of
tutorials using linux-likes, remote_console is widely documented as the
way to go, and is unavailable on windows.
This is hella confusing.
So to work around that, this patch adds an alias for 'attach' on windows
to be 'remote_console', bridging the gap.
Also the functionality was flat out broken because it would not use a
node hostname when connecting out. Since the latest release added that
functionality, this patch also fixes attach to work in the first place.
|
|
Some tool calls were outdated, and in some cases, relied on non-existing
features there.
|
|
Use either optional compilation or version-safe variants of the string
functions. Prevents warnings when the switch to OTP-21 will happen.
|
|
|
|
both double quotes and a space character
|
|
The runner script that ships with rebar builds HEART_COMMAND and run_erl arguments that preserve additional arguments the user may have passed:
https://github.com/rebar/rebar/blob/master/priv/templates/simplenode.runner#L215-L238
This PR preserves this behavior. In additon, the current code on this line sets $@ but does not do anything with the result:
https://github.com/erlware/relx/blob/master/priv/templates/extended_bin#L481-L482
Investigated in response to this ML thread: http://erlang.org/pipermail/erlang-questions/2017-October/093974.html
|
|
|
|
Add boot_var ERTS_LIB_DIR to relx_get_nodename
|
|
* Extended start script command extensions
Provide a mechanism that allows for the application
to extend the list of commands available to be invoked
from the start script. An application may be able to
define a 'foo' extension that is associated with a 'foo_script'
written and maintained by the applicationr, (this association is
kept in rebar.config), upon invocation of bin/<release_name> foo
the 'foo_script' will then be invoked.
* Add test coverage for extension script
* Ensure extended script usage argument
|
|
|
|
[windows] pass extra arguments from vm.args to service
|
|
vm.args check for name/sname parameter
|
|
- distinction between non-existing and existing but non-readable args_file
- fixed circularity check to include the base vm.args file
|
|
vm.args and referenced args_files will now be checked for:
- non-existing -args_files
- circular dependencies between -args_files
- relative paths in -args_files
- multiple/mixed occurences of -name and -sname parameters
- missing -name or -sname parameters
|
|
|
|
|
|
usage of \s in awk regexp is a gawk extension. OSX comes with a different variant of awk. This fix should make the awk code POSIX-compliant and should work in all variants of awk (tested with gawk --traditional)
|
|
The current version of extended_bin checks if there is a name or sname parameter in vm.args and refuses to start if there is none. However, it is allowed that the vm.args file (more abstract, any -args_file that is given to erl/erlexec etc) itself may contain -args_file parameters (see http://erlang.org/doc/man/erl.html), which may contain the name/sname parameters.
This change will recursively scan the files mentioned in -args_file parameters in vm.args as well as -args_file parameters in the mentioned files etcetc, and return the first occurence of a name/sname parameter.
Two points are worth mentioning, though:
- The name/sname check works only with absolute paths in the args_file parameters. Relative paths are probably a bad idea there, anyway, since it would make any setup rather fragile.
- There is no check for circular dependencies. There was none before, and this change does not add any.
|
|
|
|
|
|
|
|
As mentioned in the (closed) issue #127, the default pipe directory is not multi-user friendly. When running multiple relx-generated releases as different users, the first to be started will create the /tmp/erl_pipes directory, which the ones starting after that cannot create their pipe dirs in. The ticket mentions that this should be fixed by #154, but I cannot see how this is supposed to solve the issue. Right now, the only way I see is to use prefix the start command with env PIPE_DIR=/some/other/dir/, which is quite cumbersome.
This change will work exactly as before when PIPE_DIR is explicitly set. In case PIPE_DIR is not set, however, it will first attempt to create the parent dir /tmp/erl_pipes with mode 1777, which should enable releases being run as other users to create their pipe dirs.
|
|
|
|
|
|
Any output string from .erlang to stdout will be prepend to
code:root_dir(), so it will cause the retrieved dir incorrect.
The fix is to start erl with the no_dot_erlang boot file.
|
|
|
|
|
|
The syntax for proper path handling when the path may contain
parentheses (such as C:\Program Files (x86)\...) requires quoting the
entire assignment expression:
set "var=%val%"
any other way of doing it may interpret the parentheses and cause
failures.
|
|
be 'install' or 'unpack'
|
|
Provide a status command to start script which,
by default, runs a builtin hook that simply prints
which applications are running in the node.
This hook can then be customized to print whatever
the user wants by adding
{status, [{custom, "path/to/hook"}]}
to already existing extended_start_script_hooks.
|
|
|
|
|
|
Fix OSX readlink
|
|
It does not support the -f option.
|
|
When replacing OS vars only create multi-node sys.config
files (ie. filenames containing the node name) only if explicitly
requested. This is necessary because inclusion of other config
files inside of sys.config is only allowed if the file is actually called
sys.config and nothing else.
|
|
|
|
Allow upgrade/downgrade from non-permanent versions
|
|
New 'extended_start_script_hooks' directive that allows the
developer to define six different hook shell scripts to be invoked
at pre/post start/stop/install upgrade phases.
Besides these custom defined scripts, other types of builtin
scripts are also available, these offer pre-packaged functionality
that can be used directly, they are:
pid - writes the beam pid to a configurable file location
(/var/run/<rel_name>.pid by default).
wait_for_vm_start - waits for the vm to start (ie. when it responds to pings)
wait_for_process - waits for a configurable name
to appear in the erlang process registry
The hook scripts are invoked with the 'source' command, therefore
they have access to all the variables in the start script.
|
|
Now that you can upgrade to a new version
without setting it as permanent you should also
be able to downgrade from it, this allows downgrading
or upgrading to a version that is already marked as
permanent but is not actually currently running.
|
|
Move it next to console since it's mostly the
same thing but with extra parameters.
|
|
|
|
Support a command line argument to the `upgrade`/`downgrade`
commands: `--no-permanent`, if this is set the release will be unpacked,
installed and be made current but not permanent, when the user
is confortable with the outcomehe can issue the same command a
second time without the `--no-permanent` option.
|
|
|
|
|
|
|
|
|
|
Instead of forcing the user to put the tarball package
with the expected name (<relname>.tar.gz)and in the
expected location (releases/<version>) symlink this
fixed file name to a tarball existing in one of three different
places (releases/, releases/<version>,
releases/<version>/<relname>.tar.gz).
Refactor the install/upgrade escript to make it more
dynamic, it now runs commands that are passed from
the start script while accepting a variable number
of arguments.
Add a `versions` command to the extended start
script that prints out the currently installed versions
and their status.
|