| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
LuaRocks generally only checked whether to use a mirror in the first
operation, when it fetches the manifest. If the manifest fails to load, it
switches to the mirror and everything works from there. But if the manifest
fetches ok and the then actual rock download fails with a 504, it gives up,
instead of trying that in a mirror as well. Changing that to make it retry
every download on a mirror when the base URL matches one configured in
cfg.rocks_servers should make it much more resilient.
Fixes #1299.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
* Fix. Do not revoke permission from the current user
* Do not quote FS commands
* hotfix: remove stray character
Co-authored-by: Hisham Muhammad <hisham@gobolinux.org>
|
| |
|
|
|
|
|
|
| |
For environments other than UNIX, this will return false.
Suggested by @hishamhm.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The loading of the config file has been changed to support the XDG
Base Directory specification. More info here:
https://specifications.freedesktop.org/basedir-spec/basedir-spec-latest.html
If in a UNIX environment, LuaRocks will try loading a config file in
XDG_CONFIG_HOME. If this fails, try loading one in the old location
~/.luarocks.
This method clashed a bit with the old code that separated the setting
of confdirs and the loading of the config files. As we now change
cfg.homeconfdir depending on where we find the config file I found it
necessary to remove the function set_confdirs() and move some pieces
into the config loading logic.
|
|
|
|
|
|
|
| |
This is needed in order to follow the XDG Base Directory specification.
More info here:
https://specifications.freedesktop.org/basedir-spec/basedir-spec-latest.html
|
| |
|
|
|
|
|
|
|
|
|
| |
Don't set cfg.home_tree to nil in cfg.lua when the user is root.
Instead, do explicit checks if the user is root in cmd.lua.
This removes some confusion, as cfg.home_tree is later set by the
default system config file, so it will show up when running 'luarocks
config'.
|
|
|
|
|
|
| |
https://github.com/luarocks/luarocks/issues/1293
Co-authored-by: Hisham Muhammad <hisham@gobolinux.org>
|
|\ |
|
|/ |
|
|
|
|
| |
luaposix wasn't updated for Lua 5.4 yet
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
Older versions of LuaRocks (< 3) registered "foo.init" files as "foo"
(which caused problems, so that behavior was changed). But look for that
in the manifest anyway for backward compatibility.
See https://github.com/luarocks/luarocks/commit/65c417e0ecda55f44c691df032163a8c08f0b52a
See #1276.
|
|
|
|
|
|
|
| |
Ensure that `luarocks` fails if an installation does not successfully
deploy all files registered in the `rock_manifest`.
See #1276.
|
|
|
|
|
|
|
| |
If rock_manifest could not be found, the entry in manifest is unusable:
without the list of files from rock_manifest, we can't scan the repository
to remove files. This means the entry in the repo manifest is a leftover from
an incomplete removal. Remove the entry from the repo manifest.
|
| |
|
|
|
|
|
|
|
| |
* Add fs abstraction file for NetBSD
* Add init function with verification that the unzip binary is
available
Co-authored-by: Lawrence Hoffman <lawrence.hoffman@block-g.com>
|
| |
|
| |
|
| |
|
|
|
|
|
| |
Old versions of LuaRocks might get confused with modules called "x.init".
This makes the manifest writer more resilient to that.
|
|
|
|
| |
This is to match the behavior of luarocks.deps
|
|
|
|
| |
closes #1280
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
Tried to do this as non-intrusively as possible, mostly by ignoring the
issues in-line.
Set the option unused_secondaries to false as suggested by @hishamhm.
This makes named but ununsed return values, that serves a documenting
purpose, allowed.
|
|
|
|
| |
I realized I had this around for ages but never committed it.
|
| |
|
| |
|
| |
|
|
|
|
|
| |
Add a changelog for version 3.5, summarizing the announcement posted to
the lua-l mailing list.
|
|
|
|
| |
Fixes #1238.
|
|
|
| |
Reference: https://svnweb.freebsd.org/ports/head/devel/lua-luarocks/files/patch-src_luarocks_core_cfg.lua?revision=477367&view=markup (based on)
|
|\ |
|
|/ |
|
|
|
|
|
|
|
| |
Reverts 04513372d7d5859ea865252095c3b5ce0762228f
Fixes #1232.
Fixes #1233.
|
|
|
|
|
|
|
| |
Previously, if LUA_INCDIR was specified but invalid, install and other
commands would error without a message, eventually causing a failed
assertion in cmd.lua and leaving a hard-to-troubleshoot situation
with no information.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Currently, LuaRocks supports:
* (a) Lua interpreters built for MSYS2 (Lua interpreters depend on
msys-2.0.dll).
(the "msys" platform)
* (b) Lua interpreters built by MinGW (Lua interpreters don't depend
on msys-2.0.dll).
(the "mingw" platform)
This change adds support for (c) Lua interpreters built as native
Windows application by MSYS2 + Mingw-w64 (Lua interpreters don't
depend on msys-2.0.dll). (the "msys2_mingw_w64" platform)
Here are differences between (a), (b) and (c):
* (a) can't work without MSYS2 (msys-2.0.dll)
* (b) can work without MSYS2
* (c) can work without MSYS2 but is generally used with MSYS2
because MSYS2 provides packages of useful libraries such as libxml2.
This change assumes that users use (c) with MSYS2. But this change
still uses win32/tools provided by LuaRocks not MSYS2.
MSYS2 has LuaRocks package:
https://github.com/msys2/MINGW-packages/tree/master/mingw-w64-lua-luarocks
It applies a patch to support (c). If this change is merged into
LuaRocks, MSYS2 doesn't need to have a patch for LuaRocks.
|
|
|
|
|
| |
* fs: fix verbose mode when an argument is nil
* Use table.pack in more places
|
| |
|
| |
|
|\ |
|
|/ |
|