This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: setup.exe suggestion + patch


On Thu, 6 Sep 2007, Lewis Hyatt wrote:

Hello-

Firstly, thanks to everyone who has worked on setup.exe, it's really a
very convenient program! There is just one thing that has always
bothered me, which is that you have to click repeatedly on the package
or category to cycle through all the available actions to find the one
you want. The main problem is that each click causes the dependencies to
be recalculated, which can cause annoying slowdowns if you're trying to
do something like uninstall all packages in a large category. There is
also the following situation which occurs often, especially when you are
playing around with installing and uninstalling new packages:

-package A requires package B
-package A has two available versions
-package B appears before package A in the list

Now suppose A and B are both installed, and you want to uninstall them.
Since B appears first, you click through to uninstall, no problem. Now
you scroll down, maybe several pages away, and try to uninstall package
A. The first time you click, though, you end up on the Prev version,
which then calculates that it needs package A and goes back and sets
package A to Install again. The only way to uninstall both of them is to
uninstall B first, and then A. When there are multiple dependencies
involved, it can quickly get impossible to get setup to do what you
want.

The simplest way I could think of to correct this would be to change the
behavior so that when you click on a Category or a Package, instead of
simply cycling through, you get a little popup menu that asks you what
you want to do instead. This way, you can go directly to Uninstall
without dealing with the intervening options. This also lets you see all
available versions at once, and avoids calculating dependencies
unnecessarily.

I wrote a simple patch that implements this suggestion. Attached are the
outputs of cvs diff (in diff.txt) and cvs diff -n (in diff_n.txt). (I'm
sorry, I don't know much about CVS, is this the preferred way to submit
a patch?). Here is a summary of the changes:

-Created new class PopupMenu in PopupMenu.{h,cc}, which makes a popup
menu at the mouse cursor location and returns the selected item.

-Added #define to resource.h for use by PopupMenu. For now, it just
reserves 100 IDs, supporting arbitrary popup menus with up to 100
entries. (The number 100 is easily configurable in resource.h.)

-Modified PickCategoryLine to open the menu instead of cycling.

-Added new function select_action() to the packagemeta class, which
implements the menu selection. For now, this is done in an extremely
quick and dirty way that simple calls set_action() repeatedly to figure
out which options would have been cycled through. I would be willing to
re-do this in a more efficient way if this patch is deemed useful, but I
don't even think that's necessary, I think it's fine to do it this way
which reuses the already bug-tested code in set_action().

-Modified PickPackageLine to call select_action() instead of
set_action() when the line is clicked.

-Made some minor changes to packagemeta::_action to expose the category
strings as part of the public interface, so they could be reused in the
popup menu.

Anyway I hope this is useful, if this patch isn't acceptable please let
me know and I can fix it or change it. I wasn't sure about conventions
with tabs, line endings, line lengths, etc., for one thing. In general,
I think the problem I have described requires fixing. If you don't think
this solution is an improvement, I can look into fixing it a different
way also.

First off, thank you for the popup menu implementation -- I was planning to use it for something else in setup, and now I don't have to write it.

Second, I agree that the problem is good, and I like your solution of
selection vs. cycling.

Third, I have to apologize -- I've had a partial reply to your message
sitting in my drafts since the day you sent it, but got bogged down.

A few comments on the patch:
1) It would be great if you used "diff -up" -- unified diffs are so much
easier to read.
2) Is there a reason you use popup menus, rather than pull-down lists?
3) [Minor] You'd use the GNU coding guidelines for whitespace and
indentation.

If you could resend a unified diff, I'll apply it in my tree and test it
out.
	Igor
--
				http://cs.nyu.edu/~pechtcha/
     |\      _,,,---,,_	    pechtcha@cs.nyu.edu | igor@watson.ibm.com
ZZZzz /,`.-'`'    -.  ;-;;,_		Igor Peshansky, Ph.D. (name changed!)
    |,4-  ) )-,_. ,\ (  `'-'		old name: Igor Pechtchanski
   '---''(_/--'  `-'\_) fL	a.k.a JaguaR-R-R-r-r-r-.-.-.  Meow!

Belief can be manipulated. Only knowledge is dangerous. -- Frank Herbert

--
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/


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]