Corinna Vinschen wrote:
> This (untested) patch to the service helper script
> /usr/share/csih/cygwin-service-installation-helper.sh
> should fix that problem.
>
>
> Index: cygwin-service-installation-helper.sh
> ===================================================================
> RCS file:
> /sourceware/projects/cygwin-apps-home/cvsfiles/csih/cygwin-service-installation-helper.sh,v
> retrieving revision 1.15
> diff -u -p -r1.15 cygwin-service-installation-helper.sh
> --- cygwin-service-installation-helper.sh 22 Aug 2008 03:56:43 -0000
> 1.15
> +++ cygwin-service-installation-helper.sh 8 Dec 2008 15:10:36 -0000
> @@ -2277,7 +2277,7 @@ csih_select_privileged_username()
> echo ""
> csih_inform "This script plans to use '${username}'."
> csih_inform "'${username}' will only be used by registered services."
> - if csih_request "Do you want to use a different name?"
> + if ! csih_request "Do you want to use this account name for running
> this service?"
> then
> csih_get_value "Enter the new user name:"
> username="${csih_value}"
As I've pointed out before, this breaks csih clients that are run as
part of postinstall. Now, I had thought that there were more of those
in the past, than the paltry example below.
inetutils postinstall:
export IU_CONFIG_AUTO_ANSWER_NO=1
/usr/bin/iu-config
export SYSLOGD_CONFIG_AUTO_ANSWER_NO=1
/usr/bin/syslogd-config
You /REALLY/ don't want the postinstall script to run with
AUTO_ANSWER=yes. Hence, the sense of the question is set up so that no
user input is ever required when AUTO_ANSWER=no.
However, no such guarantee is made when AUTO_ANSWER=yes.
I don't see how to accomodate both requirements, so that --yes "does the
right thing" and --no "does the right thing" for a question phrased
identically. Perhaps that question, specifically, should be handled
specially? (Pseudo-code):
if auto_answer=yes
if ! csih_request "Do you want to use this account name for running
this service"
# since auto-answer=yes, we never get here, but we "ask" the
# question anyway so that it gets printed out using the 'question'
# colorization
:
...
elif auto_answer=no
if csih_request "Do you want to use a different name?"
# since auto-answer=no, we never get here, but we "ask" the
# question anyway so that it gets printed out using the 'question'
# colorization
...
else
if csih_request "Do you want to use a different name?"
csih_get_value "Enter the new user name:"
username="${csih_value}"
...
...
But that would just confuse someone who came at this "cold": look at the
manual transcript of a foo-install session, and then said, OK, I want to
automate this. Should I use --yes, or --no? With the pseudo-code above,
the answer (as far as the manual transcript would indicate) would be
--yes, except I think I want --no for "Do you want to use a different
name?" (e.g. I want the default prived user).
Okay. So, suppose we modify the pseudo-code and word the non-auto-answer
question like the auto-answer=yes case. Then, somebody looking at a
manual transcript might ask: I want to run this script "safely" so that
things are set up, but no services are actually installed and no new
user accounts are created. Should I use --yes, or --no? Under this
modified pseudocode, the answer would be --no, except I think I want
--yes (e.g. I want the default prived user) for "Do you want to use this
account name for running this service"
Anybody have a better idea, that works for both "automated" cases yet
doesn't confuse the heck out of somebody trying to figure out which of
--yes or --no to use when implementing their own automation?
--
Chuck
--
Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
Problem reports: http://cygwin.com/problems.html
Documentation: http://cygwin.com/docs.html
FAQ: http://cygwin.com/faq/
