61348 - Create a mathml.xpi

Status: VERIFIED WONTFIX

(Core :: MathML, defect, P3)

Description

[Axel Hecht]

I file this bug, to track some work (and errors) while creating a mathml.xpi
for mozilla 0.6

I generated a patch to create a mathml.xpi, zipping together
$(DIST)/bin/components/libucvmath.so 
$(DIST)/bin/components/libgklayout.so 
$(DIST)/bin/components/libgfx_gtk.so 

These include all code with MOZ_MATHML as far as I found it.
I created a install script as well.

To test this I created two optimized stripped non-debug builds, one with
ac_add_options --with-extensions=all,transformiix,cview
ac_add_options --enable-mathml
ac_add_options --enable-svg
ac_add_options --enable-xsl
ac_add_options --enable-mng
ac_add_options --enable-ldap
the other with
ac_add_options --enable-mng

I created the xpi from the first build and installed it in the second. All three
libs made it to bin/components, and a regExport|grep -i math yields the same
results for both the mathml build as for the installed one.

But only the mathml build will render mathml, the one with the installed ones
doesn't do anything.

Anything more I could test?

Axel

PS: I'll attach the patch to layout/mathml/Makefile.in and the mathml.jst in a 
minute

Comment 1

[Axel Hecht]

Attachment: the mathml.jst file, builds install.js

Comment 2

[Axel Hecht]

Attachment: patch to mozilla/layout/mathml/Makefile.in, try make xpi

Comment 3

[Philip Langdale]

You will also need to include:

$(DIST)/bin/dtd/mathml.dtd
$(DIST)/bin/res/mathml.css

There may be others, I'm not 100% sure.

Comment 4

[Dawn Endico]

I bet your problem is that you turned on lots of options but are only
including libs needed by mathml. Lets do this with mathml as the only
added option. At least until we prove that this idea of replacing
layout libraries is going to work.

Also, note that rbs is away till january.

Comment 5

[Axel Hecht]

Philip, thanx, you got it.

Now we have just one problem left:

The during the install phase of the build, layout/mathml/content/src/Makefile.in
runs
$(PERL) $(srcdir)/mathml-css.pl $(DIST)/bin/res/ua.css
which adds mathml.css to ua.css.
should we ship a modified ua.css (bad, IMHO), or should I try to get something
working in js to do the job, and add that to mathml.jst?

baseline: the current xpi works, but display is not 100% compatible without the
modification of the ua.css

patch coming

Axel

Comment 6

[Axel Hecht]

Attachment: updated patch to layout/mathml/Makefile.in to include mathml.[dtd|css]

Comment 7

[Axel Hecht]

attinasi, would you have a suggestion on the problem how to deal with a modified
ua.css for the mathml.xpi?
If not, any idea who would?

Thanx

Axel

Comment 8

[Marc Attinasi]

I'd suggest installing a new us.css that @imports the mathml.css file. ua.css
only contains @iports and two declarations, and it should be relatively stable.

I cannot think of any other way to get your style rules into the mix, except for
possibly adding C++ code to load it seperately from ua.css, possibly where
ua.css is loaded, if MATHML is defined. email me if you need more information on
this option...

Comment 9

[Axel Hecht]

Hi Robert,
added you to the CC of this bug, as you are looking into this.

I just got mail from attinasi:
<quote>
The only people who mess with ua.css are the Style Guys (including 
myself and Pierre) and the Chrome guys (mostly Hyatt). Since it really 
only contains pointers to other css files, however, it is rarely modified.

That said, I would worry that you might have to keep track of changes 
that are made to it and keep your copy in synch, and that seems like a 
pain. If the js modifications are not really difficult, I'd probably try 
that first. If it takes more than a few hours to get it right, then I'd 
do the replacement.

Another possible idea is to have an import to the mathml.css file in 
there always, and simply rely on the fact that the css file won't be 
there in non-mathml builds. If it is not there, the import will simply 
fail and the rest of the file will be left intact. If it is there, then 
it will be imported as you'd expect. This might be the easiest approach: 
1 ua.css file that never needs to be modified. Just make sure my 
assumptions about it working fine with the file missing are valid first.
</quote>

As we have to live with the ua.css being there without mathml.css (NS6, moz0.6),
I will try to look into some js to add the import statement.

But not today. So Robert, if you wanna beat me to it, you're welcome.
If the js won't go smooth, we should just replace the old ua.css, and perhaps
add a warning to the release notes of the xpi.

Axel

Comment 10

[Dawn Endico]

Instead of js, why don't you just have the installer for ns6 replace the
ua.css file? This is a one time only thing anyway since future installers
will be working with ua.css files that include mathml.css

Comment 11

[Axel Hecht]

Attachment: putting ua.css in xpi

Comment 12

[Axel Hecht]

Just attached a updated patch to layout/mathml/Makefile.in to include ua.css.
Sorry for this taking so long, I gave up on the "nice" solution to put up some
js, as there is too much stuff (it's that kind of season) on my list.
Any takers for windows/mac?

Axel

Comment 13

[Daniel Wang]

reporter (Axel Hecht), do we still need this bug open?

Comment 14

[Axel Hecht]

I'd guess theoretically this bug would still be a nice-to-have.
In practice, shipping a mathml layout-content lib is just not feasable, so folks
working on builds that have mathml switched off (read embedding) are not going
to go thru that. As mathml is on by default, I WONTFIX this.

Comment 15

[Daniel Wang]

v