Bugzilla Bug 182855

For users who want to test the MSIE7 installation (it is still in alpha), here
is an ebuild for IEs4Linux-2.5_beta6.

Created an attachment (id=122764) [edit]
ies4linux-2.5_beta6.ebuild

Notes:

- When the final version of IEs4Linux-2.5 will be released, "${MY_P}" will have
to be switched to "${P}".

- Note the updated "${HOMEPAGE}" (see
http://www.tatanka.com.br/ies4linux/news/44).

- A patch will follow, for the package.

Created an attachment (id=122765) [edit]
ies4linux-2.5_beta6-Installation_directory_discovery.patch

This patch replaces the currently-used ies4linux-2.0.5-Adapt_paths.patch. Most
problems addressed by this previous patch, have since been addressed by
upstream. Left is only the installation directory discovery routine (as we use
a symlink, from /usr/bin/ies4linux, to the real IEs4Linux main script, we have
to handle symlinks).

Created an attachment (id=122769) [edit]
metadata.xml (updated descriptions)

Updated metadata.xml descriptions (GUI installer, MSIE 7 in alpha), in English
and French.

Depending on the MSIE7 status, in the final release of IEs4Linux, it will
possibly need to be updated again.

I forgot to add it to the ebuild, but there should probably be some RDEPEND to
dev-python/pygtk (with Python 2.4 or 2.5 -see
http://www.tatanka.com.br/ies4linux/news/41#comment-3971) and
kde-base/kommander, with some USE flags, for the graphical installers... (or
maybe just some einfo, after the installation).

If so, please update the ebuild in attachment.

Created an attachment (id=122833) [edit]
ies4linux-2.5_beta6.ebuild (updated)

Changes:

- Added "gtk" and "kde" USE flags, to trigger RDEPEND checks for PyGTK and
Kommander.

- Added some `einfo` about the IEs4Linux options to use the PyGTK and Kommander
installer interfaces.

- (Some minor coding style modifications.)

Thank you. There are still some minor things that you can change:

1. replace all einfo with elog
2. drop this `doins && doins && doins || die` scheme

otherwise this ebuild looks good for me... just make sure that there is no need
to make the unversioned dependencies (kommander, pygtk) more restrictive (e.g.
by adding version lower bound)

(In reply to comment #7)
> 
> 1. replace all einfo with elog
>


Ok (I just learnt about it, and the new, more limited meaning, of `einfo`;
thanks for the tip :)).


>
> 2. drop this `doins && doins && doins || die` scheme
> 


You mean completely removing the "|| die" for all `doins`? While `install` will
display a local error, in case of problem, it seems easy to miss (well, the
IEs4Linux emerge log is short, but still -and we never know what could happen
on a user installation).


>
> otherwise this ebuild looks good for me... just make sure that
> there is no need to make the unversioned dependencies (kommander, 
> pygtk) more restrictive (e.g. by adding version lower bound)
> 


I emailed the IEs4Linux author about it, I'll report when he will reply.

I tested with pygtk-2.6.1 and pygtk-2.8.6 (with python-2.4.4), and it seems to
work, but there is some slowness, during the installation (it blocks multiple
times, for dozens of seconds). There is no such problem with pygtk-2.10.3 and
2.10.4 (well, maybe there was some things to recompile/downgrade, after
downgrading pygtk to 2.6.1 or 2.8.6...).

As for Kommander, I only tested with kommander-3.5.7, and it works with this
version.

> >
> > 2. drop this `doins && doins && doins || die` scheme
> > 
> 
> 
> You mean completely removing the "|| die" for all `doins`? While `install` will
> display a local error, in case of problem, it seems easy to miss (well, the
> IEs4Linux emerge log is short, but still -and we never know what could happen
> on a user installation).

Nah, your a && b && c || d notation is ambiguous. Does it mean (a && b && c) ||
d or a && (b && c || d) ? I hope you see a BIG difference between those two ;-)
The rule is: the simplier (even to read), the better.

> I emailed the IEs4Linux author about it, I'll report when he will reply.
> 
> I tested with pygtk-2.6.1 and pygtk-2.8.6 (with python-2.4.4), and it seems to
> work, but there is some slowness, during the installation (it blocks multiple
> times, for dozens of seconds). There is no such problem with pygtk-2.10.3 and
> 2.10.4 (well, maybe there was some things to recompile/downgrade, after
> downgrading pygtk to 2.6.1 or 2.8.6...).
> 
> As for Kommander, I only tested with kommander-3.5.7, and it works with this
> version.

Sounds good.

In the meantime, remove the ambiguity mentioned above and I'll add it to the
tree p.masked for testing.

Created an attachment (id=122844) [edit]
ies4linux-2.5_beta6.ebuild (updated)

Changes:

- Added some parentheses, for clarity.

- Added some "|| die", in src_unpack(), and corrected the "|| die" for the
`dodoc` (I had forgotten the `die` call).

- Replaced `einfo` with `elog`, as it is more appropriate, in this case.


(In reply to comment #9)
>
> Nah, your a && b && c || d notation is ambiguous.
> Does it mean (a && b && c) ||  d or a && (b && c || d) ?
> I hope you see a BIG difference between those two ;-)
> 


I sure do, but they are simple expressions, and this is Bash scripting :) Well,
anyway, I added parentheses, as appropriate.

Ok, I fixed your formatting a bit and commited the ebuild to the tree, thanks!

(In reply to comment #11)
> Ok, I fixed your formatting a bit and commited the ebuild to the tree, thanks!
> 

Great :)

What about the "metadata.xml" file?

Forgot about that one. Done, thanks.