1- kill -SIGUSR1
2- if there are remaining pids, wait 3 secs to get harder
3- kill -SIGTERM
4- if there are remaining pids, wait 3 secs to kill 'em all
5- kill -SIGKILL
However, it has a precondition that developers MUST respect:
An option CAN'T have differente meanings/behaviour in different subcommands.
For example, "-s" means "size" and accept an argument. If you are tempted to add
an option "-s" (that means, for example "silent", and doesn't accept an argument)
This, however, make sense even from an usability point of view.
Also, option with optional argument are prohibited.
The technique is a "double scan": first, we scan options passing
_every_ possible subcommand options to zparseopts.
This way, we can distinguish between option argument and parameters.
So, we recognize which is the subcommand.
Then, parse again
It was putting the option BEFORE the subcommand.
However, the new method supports this only partially:
* Boolean options can be put before the subcommand
* Subcommand-specific options that take an argument can be put before the
subcommand, only in the
form -kvalue ; using -k value gives an error
* Global options can always be put before the subcommand
Therefore, we'll say that options MUST go after the subcommand (to stay sure)
* parameters can be added where you want
* it's possible to use long options
* you can have "synonim" options
* it has potential for a recursive implementation
and notice if tomb hasn't right permissions
in future we should check permissions of tomb from the suid commandline
this way we can also change its permissions and make it ours
now the internal command tomb translate
generates a list of strings used inside tomb ready for translation
(as a .po file)
help needed: translations and autotools integration
exposed as command the implementation by Anathema: kills all
processes using the tomb (denying it to be closed) and umounts
tomb (the commandline script) is not supposed to be interactive
we leave interactivity to be implemented by wrappers as tomb-open
so it should stop to ask the used if to slam, rather than offer
it as a command. now also tomb-status has the menu option to slam.
piantava con cpu sparata al 100%. Non capiva su quale TTY si
trovava. Per farla breve, necessitava di sti due parametri:
OPTION ttyname=$TTY
OPTION lc-ctype=$LANG
BELLA LI'
we require users to have the mount targets for hooks already existing
this behaviour is less intrusive in someone's home that is unconscious of hooks
plus it now supports mount bind of normal files
rename your keys !
this change was motivated by the fact gpg tends to overwrite the
.tomb file if we decrypt the tomb.gpg by hand. changing the extension
in .tomb.key this doesn't happens.
tomb shell command doesn't depends from X,
all GUI notifications are moved into tomb-open including USB key handling,
gksu has been dropped completely as a method to gain privileges,
there is a new -n flag to avoid processing hooks,
variable names have been sanitized, duplicate code eliminated,
documentation has been updated and the code cleaned up
we are very close to the 1.0 now
using pinentry (with Assuan protocol) instead of our own askpass
a bit less cooler but much more secure.
this also includes partial normalization of variable names
and the redirection of tomb operational output to stderr.
Now, the post-hooks is passed an argument (open or close) that can be
used to launch commands when the tomb opens or closes, respectively.
This patch also introduces a safer version of bind-hooks that doesn,t
eval the contents, but read a map:
local/to/tomb local/to/HOME
The local/to/HOME is hardly enforced ATM.
just from the tomb commandline for now (we start missing a simple GUI wizard eh)
also includes fixes for backward compat (renames tomb volumes after files)
this helps scripts track all mounted tombs via mount -l which prints labels
and some tomb-open wrapper fixes too
