X-Git-Url: https://git.cworth.org/git?p=obsolete%2Fnotmuch-wiki;a=blobdiff_plain;f=remoteusage%2Faboriginal.mdwn;h=e69de29bb2d1d6434b8b29ae775ad8c2e48c5391;hp=3d012b8eed684b115b2f2adc9b5820ff52d7831e;hb=d4cc0b096cdfd6526c290437fdb43b2953a2f78d;hpb=61f624d2aa54c56405d47e5a5454622639638494 diff --git a/remoteusage/aboriginal.mdwn b/remoteusage/aboriginal.mdwn index 3d012b8..e69de29 100644 --- a/remoteusage/aboriginal.mdwn +++ b/remoteusage/aboriginal.mdwn @@ -1,155 +0,0 @@ -## Remoteusage without password-free login requirement - -This is alternative to [[remoteusage|remoteusage]] where password-free -login is not a requirement. See [[remoteusage|remoteusage]] page for -other requirements and general information. - -This solution uses one pre-made ssh connection where the client is put -into "master" mode (-M) for connection sharing. The wrapper script then -uses the control socket created by this pre-made ssh connection for -its own connection. As long as master ssh connection is live, slave -can use it. Disconnecting master all future attempts to connect -from the script will fail. - -At the end of this document there is information for some possible ways -how master ssh connection can be done. - -## The script - -Write the following code to a file, for example `remote-notmuch.sh`. - - #!/bin/bash - - # http://notmuchmail.org/remoteusage/aboriginal/ - - set -eu - # To trace execution, uncomment next line. - #BASH_XTRACEFD=6; exec 6>>remote-errors; echo -- >&6; set -x - - readonly SSH_CONTROL_SOCK='~'/.ssh/master-user@host:22 - - readonly notmuch=notmuch - - printf -v ARGS '%q ' "$@" # bash feature - - readonly SSH_CONTROL_ARGS='-oControlMaster=no -S '$SSH_CONTROL_SOCK - - if ssh -q $SSH_CONTROL_ARGS 0.1 $notmuch $ARGS - then exit 0 - else ev=$? - fi - - # continuing here in case ssh exited with nonzero value. - - case $* in - 'config get user.primary_email') echo 'nobody@nowhere.invalid'; exit 0 ;; - 'config get user.name') echo 'nobody'; exit 0 ;; - 'count'*'--batch'*) while read line; do echo 1; done; exit 0 ;; - 'count'*) echo 1; exit 0 ;; - 'search-tags'*) echo 'errors'; exit 0 ;; - 'search'*'--output=tags'*) echo 'errors'; exit 0 ;; - esac - - if ssh $SSH_CONTROL_ARGS -O check 0.1 - then - echo ' Control socket is alive but something failed during data transmission.' - exit $ev - fi - - echo " See`sed '1d;2d;s/.//;q' "$0"` for help." - #EOF - -Note the `0.1` in ssh command line. It is used to avoid any opportunistic -behaviour ssh might do; for example if control socket is not alive ssh -would attempt to do it's own ssh connection to remote ssh server. As -address `0.1` is invalid this attempt will fail early. - -## Test - -Easiest way to test this script is to run the pre-made ssh connection -using the following command line: - - ssh -M -S '~'/.ssh/master-user@host:22 [user@]remotehost sleep 600 - -(replace `[user@]remotehost` with your login info). Doing this the -above wrapper script can be run unmodified. After the above command has -been run on **one terminal**, enter `chmod +x remote-notmuch.sh` in -**another terminal** and then test the script with - - ./remote-notmuch.sh help - -Note that the '~' in the ssh command line above is inside single quotes -for a reason. In this case shell never expand it to `$HOME` -- ssh does -it by not reading `$HOME` but checking the real user home directory -from `/etc/passwd`. For security purposes this is just how it should -be. - -## Tune - -The path `'~'/.ssh/master-user@host:22` might look too generic to be -used as is as the control socket after initial testing (but it can -be used). It is presented as a template for what could be configured -to `$HOME/.ssh/config`. For example: - - Host * - ControlPath ~/.ssh/master-%h@%p:%r - -is a good entry to be written in `$HOME/.ssh/config`; -[[remoteusage|remoteusage]] uses the same. Now, let's say you'd -make your pre-made ssh connection with command - - ssh -M alice@example.org - -After configuring -`readonly SSH_CONTROL_SOCK='~'/.ssh/master-alice@example.org:22` -to the `./remote-notmuch.sh` wrapper script testing with -`./remote-notmuch.sh help` should work fine. - -## Configure Emacs on the client computer ## - -See the section *Configure Emacs on the client computer* in -[[remoteusage|remoteusage]] how to do this. The instructions are the same. - - -## Creating master connection - -As mentioned so many times, using this solution requires one pre-made -ssh connection in "master" mode. The simplest way is to dedicate one -terminal for the connection with shell access to the remote machine: - - ssh -M -S '~'/.ssh/master-user@host:22 [user@]remotehost - -One possibility is to have this dedicated terminal in a way that the -connection has (for example 1 hour) timeout: - - ssh -M -S '~'/.ssh/master-user@host:22 [user@]remotehost sleep 3600 - -The above holds the terminal. The next alternative puts the command in -background: - - ssh -f -M -S '~'/.ssh/master-user@host:22 [user@]remotehost sleep 3600 - -If you don't want this to timeout so soon, use a longer sleep, like 99999999 -(8 9:s, 1157 days, a bit more than 3 years). - -A more "exotic" solution would be to make a shell script running on remote -machine, checking/inotifying when new mail arrives. When mail arrives it -could send message back to local host, where a graphical client (to be written) -pops up on display providing info about received mail (and exiting this -graphical client connection to remote host is terminated). - -## Troubleshooting - -If you experience strange output when using from emacs first attempt to just -run - - ./remote-notmuch.sh help - -from command line and observe output. If it looks as it should be next uncomment -the line - - #BASH_XTRACEFD=6; exec 6>>remote-errors; echo -- >&6; set -x - -in `./remote-notmuch.sh` and attempt to use it from emacs again -- and then -examine the contents of `remote-errors` in the working directory emacs was -started.