Firefox's strange behavior

classic Classic list List threaded Threaded
19 messages Options
Reply | Threaded
Open this post in threaded view
|

Firefox's strange behavior

Einar Lygre-2
Firefox is a great problem for me. I normally use some time to browse
through some Norwegian newspaper. Using one hour and having 10 crashes
is normal. I open 7 newspapers in 7  tabs and already during the opening
process FF often crashes. Restarting and the tabs are loaded. If I
select a link from within a website FF very often crashes. Then it may
help (for a time) to close other tabs and just use one. Some crashes
also freezes the machine, and only a very few gives any error messages.
Restarting FF after some of the crashes gives an apologize, other start
from the point of crash.
Now a new problem has surfaced. In two of the newspapers (www.ap.no and
www.bt.no) I can't use the right mouse button on a link to get up the
menu to open the new site in a new tab.  The menu doesn't show up. These
two newspapers are recently bought up by a company, and I guess they now
are using some common software to create the sites. But this is not a
problem in MS Vista.
I should like to know if all these crashes is a common problem, or if it
is just how it is in eCS. I think I have done all the right things and
updated to latest version. New profile, new fontconfig.
FF   10.0.12,  eCS 2.1

Einar Lygre
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Paavo Nurminen-2
Einar Lygre wrote:

> Firefox is a great problem for me. I normally use some time to browse
> through some Norwegian newspaper. Using one hour and having 10 crashes
> is normal. I open 7 newspapers in 7 tabs and already during the opening
> process FF often crashes. Restarting and the tabs are loaded. If I
> select a link from within a website FF very often crashes. Then it may
> help (for a time) to close other tabs and just use one. Some crashes
> also freezes the machine, and only a very few gives any error messages.
> Restarting FF after some of the crashes gives an apologize, other start
> from the point of crash.
> Now a new problem has surfaced. In two of the newspapers (www.ap.no and
> www.bt.no) I can't use the right mouse button on a link to get up the
> menu to open the new site in a new tab. The menu doesn't show up. These
> two newspapers are recently bought up by a company, and I guess they now
> are using some common software to create the sites. But this is not a
> problem in MS Vista.
> I should like to know if all these crashes is a common problem, or if it
> is just how it is in eCS. I think I have done all the right things and
> updated to latest version. New profile, new fontconfig.
> FF 10.0.12, eCS 2.1
>
> Einar Lygre

I can see the same effect here with Aftenposten and Berlinske Tidende
as you see. But the links work, and you can move back to original.
This one works as you wished, www.yle.fi.

I do not usually use Firefox, SeaMonkey does crash, but not very often.

eCS 2.1 and SeaMonkey 2.7.2.

Paavo Nurminen

_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Andreas Schnellbacher
In reply to this post by Einar Lygre-2
Einar Lygre wrote:

> Firefox is a great problem for me. I normally use some time to
> browse through some Norwegian newspaper. Using one hour and having
> 10 crashes is normal.

I know such behavior. I'm thinking it's caused by JavaScript code. I
found out that some versions work better than others - but only in
some cases. IMO Mozilla entered the alpha stage (for the OS/2 version)
soon after Peter discontinued development. Let's hope for Dmitry's
version. From the GitHub logs, it looks like he's not that far from a
release.

--
Andreas Schnellbacher
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Andreas Schnellbacher
Andreas Schnellbacher wrote:

> Einar Lygre wrote:
>
>> Firefox is a great problem for me. I normally use some time to
>> browse through some Norwegian newspaper. Using one hour and having
>> 10 crashes is normal.
>
> [...]

A few other thoughts than those from my last posting:

In many cases a new profile cured most of the problems. Note that I
have permanent profile problems with the more recent 32 bit W$
versions at work since years. Additionally, the Mozilla Foundation
replaced with their "Rapid Release" strategy stability with release
frequency.

--
Andreas Schnellbacher
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

David T. Johnson
In reply to this post by Einar Lygre-2
Einar Lygre wrote:

> Firefox is a great problem for me. I normally use some time to browse
> through some Norwegian newspaper. Using one hour and having 10 crashes
> is normal. I open 7 newspapers in 7  tabs and already during the opening
> process FF often crashes. Restarting and the tabs are loaded. If I
> select a link from within a website FF very often crashes. Then it may
> help (for a time) to close other tabs and just use one. Some crashes
> also freezes the machine, and only a very few gives any error messages.
> Restarting FF after some of the crashes gives an apologize, other start
> from the point of crash.
> Now a new problem has surfaced. In two of the newspapers (www.ap.no and
> www.bt.no) I can't use the right mouse button on a link to get up the
> menu to open the new site in a new tab.  The menu doesn't show up. These
> two newspapers are recently bought up by a company, and I guess they now
> are using some common software to create the sites. But this is not a
> problem in MS Vista.
> I should like to know if all these crashes is a common problem, or if it
> is just how it is in eCS. I think I have done all the right things and
> updated to latest version. New profile, new fontconfig.
> FF   10.0.12,  eCS 2.1
>
> Einar Lygre

These are problems with javascript.  Firefox 10 (and the equivalent Sea
Monkey) are getting to be quite out of date and are unable to handle
properly some of the newer javascript things that these sites are doing.
    You would not see these crashes if you turn javascript off but then
the web site probably would not be usable if you did that.  :(

--
Posted with OS/2 Warp 4.52
and Sea Monkey 1.5a
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Dave Yeo-3
In reply to this post by Einar Lygre-2
Einar Lygre wrote:

> Firefox is a great problem for me. I normally use some time to browse
> through some Norwegian newspaper. Using one hour and having 10 crashes
> is normal. I open 7 newspapers in 7 tabs and already during the opening
> process FF often crashes. Restarting and the tabs are loaded. If I
> select a link from within a website FF very often crashes. Then it may
> help (for a time) to close other tabs and just use one. Some crashes
> also freezes the machine, and only a very few gives any error messages.
> Restarting FF after some of the crashes gives an apologize, other start
> from the point of crash.
> Now a new problem has surfaced. In two of the newspapers (www.ap.no and
> www.bt.no) I can't use the right mouse button on a link to get up the
> menu to open the new site in a new tab. The menu doesn't show up. These
> two newspapers are recently bought up by a company, and I guess they now
> are using some common software to create the sites. But this is not a
> problem in MS Vista.
> I should like to know if all these crashes is a common problem, or if it
> is just how it is in eCS. I think I have done all the right things and
> updated to latest version. New profile, new fontconfig.
> FF 10.0.12, eCS 2.1
>
> Einar Lygre

Did you add any extensions to the new profile? If so try starting up in
safe mode and see if you get the same crashes. Also perhaps remove Flash
from your plugin directory.
You could also install no-script and only allow the minimal java-script
to run.
Others have also cured java-script related crashes by adjusting their
virtualaddress limit downward
Dave
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Dave Yeo-3
In reply to this post by Andreas Schnellbacher
Andreas Schnellbacher wrote:

> Einar Lygre wrote:
>
>> Firefox is a great problem for me. I normally use some time to
>> browse through some Norwegian newspaper. Using one hour and having
>> 10 crashes is normal.
>
> I know such behavior. I'm thinking it's caused by JavaScript code. I
> found out that some versions work better than others - but only in
> some cases. IMO Mozilla entered the alpha stage (for the OS/2 version)
> soon after Peter discontinued development. Let's hope for Dmitry's
> version. From the GitHub logs, it looks like he's not that far from a
> release.
>

Peter had a lot of time to maintain Mozilla though he was obviously
getting tired of it towards the end. It's a massive code base that could
easily keep a developer employed full time maintaining the OS/2 port.
The other problems came with the direction that Mozilla took after 3.7
became 4. Rapid release schedule means more changes and more things to
break. Moving to xul means having a DLL that is right at the limits of
our old operating system (or perhaps the linkers). For a long time we
haven't been able to build a debug version or even a version with debug
symbols enabled. Working with Steven we finally figured out that the
problem was happening when the object size of xul.dll was over 32MBs.
Hopefully Dimitri will be able to build debug versions of 17esr as
debugging is much easier and also hoping that the IPC stuff doesn't push
the DLL over the size limit.
Without the IPC stuff versions 5 to 10 were sorta mickey moused together
and became harder and harder to build, especially for non-programmers
such Walter and me.
Dimitri seems pretty knowledgeable and has learned a lot about the
Mozilla code since he started working on it but as I said it is a huge
complex code base that really could use a few developers working full
time on it. It also really needs a more modern operating system then
OS/2, we're missing the 3d stuff, some font stuff, large continuous
chunks of memory and probably more.

Dave
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Cameron Kaiser-2
On 8/11/13 7:34 PM, Dave Yeo wrote:

> Peter had a lot of time to maintain Mozilla though he was obviously
> getting tired of it towards the end. It's a massive code base that could
> easily keep a developer employed full time maintaining the OS/2 port.
> The other problems came with the direction that Mozilla took after 3.7
> became 4. Rapid release schedule means more changes and more things to
> break. Moving to xul means having a DLL that is right at the limits of
> our old operating system (or perhaps the linkers). For a long time we
> haven't been able to build a debug version or even a version with debug
> symbols enabled. Working with Steven we finally figured out that the
> problem was happening when the object size of xul.dll was over 32MBs.
> Hopefully Dimitri will be able to build debug versions of 17esr as
> debugging is much easier and also hoping that the IPC stuff doesn't push
> the DLL over the size limit.
> Without the IPC stuff versions 5 to 10 were sorta mickey moused together
> and became harder and harder to build, especially for non-programmers
> such Walter and me.
> Dimitri seems pretty knowledgeable and has learned a lot about the
> Mozilla code since he started working on it but as I said it is a huge
> complex code base that really could use a few developers working full
> time on it. It also really needs a more modern operating system then
> OS/2, we're missing the 3d stuff, some font stuff, large continuous
> chunks of memory and probably more.

FWIW, these are the same problems we deal with in the TenFourFox port.
We don't implement IPC either (OS X 10.4 doesn't have spawn() and
friends), we don't have 3D (OpenGL 1.1 only), we don't support jemalloc
(lots of weird bugs when we try to turn it on) and we eventually had to
hack the OS X PowerPC linker to get libxul to build properly. For
Firefox 19 we moved entirely to a new, non-Apple gcc toolchain with our
hacked linker, and I do most of my work in an upgraded gdb I hacked
myself. But the browser still gets up and runs.

The point being, Firefox doesn't actually *need* these things to
function, though they do work better with them. We're fortunate in that
Harfbuzz works pretty well on 10.4, so fonts haven't been a big problem
for us other than altering gfx/thebes/ to use the old ATSUI font API
instead of the current CoreText one, so we've been able to hold those
things together and instead put effort into performance, such as a
PowerPC JavaScript JIT and AltiVec acceleration. I don't know how well
the x86 JIT works on OS/2, but I imagine it works at some level.

For OS/2, other than fonts, I think your biggest challenge will be the
build system, and not just because of the present work Gregor is doing.
If someone is able to get Python 2.7.3 to be happy on OS/2, that will go
a long way to improving the longevity of the OS/2 port. Additionally, I
think that whoever maintains Firefox OS/2 also needs to be ruthless
about removing features that can't be made to work if they are not
essential. We've jettisoned plugins entirely in TenFourFox because
they're just not workable, for example. I see you've made that choice
for downloadable fonts, which is unfortunate, but may be a better option
than fruitless struggles in the debugger.

Mozilla is starting to talk more about resurrecting Electrolysis (e10s),
for which IPC is a prerequisite. For Fx25 this is just thumbnails, so it
can be turned off (not essential). However, the talk of better
sandboxing presupposes detached processes, so I've started to look again
at implementing some of the spawn() family of functions in userspace.
OS/2 will probably need to do the same.

Cameron Kaiser
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Peter Brown-2
In reply to this post by Einar Lygre-2
Hi Einar

Einar Lygre wrote:

> Firefox is a great problem for me. I normally use some time to browse
> through some Norwegian newspaper. Using one hour and having 10 crashes
> is normal. I open 7 newspapers in 7 tabs and already during the opening
> process FF often crashes. Restarting and the tabs are loaded. If I
> select a link from within a website FF very often crashes. Then it may
> help (for a time) to close other tabs and just use one. Some crashes
> also freezes the machine, and only a very few gives any error messages.
> Restarting FF after some of the crashes gives an apologize, other start
> from the point of crash.
> Now a new problem has surfaced. In two of the newspapers (www.ap.no and
> www.bt.no) I can't use the right mouse button on a link to get up the
> menu to open the new site in a new tab. The menu doesn't show up. These
> two newspapers are recently bought up by a company, and I guess they now
> are using some common software to create the sites. But this is not a
> problem in MS Vista.
> I should like to know if all these crashes is a common problem, or if it
> is just how it is in eCS. I think I have done all the right things and
> updated to latest version. New profile, new fontconfig.
> FF 10.0.12, eCS 2.1
>
> Einar Lygre



I'm using Seamonkey 2.7.2-12 which is, I believe, based on the Firefox
10.0.12 code.

I cannot comment on the tabs problem as I rarely have more than 5 tabs
open except to say that the few times I have had more than 5 tabs open
Seamonkey has not crashed as a result. I suspect available RAM and
system configuration may have a bearing on the problem.

No problems using the 2 websites you mention; RMB on a link and the
options menu appears. Maybe I'm trying the wrong links...

I do have a useful extension installed called NoScript
http://noscript.net/ which may help with javascript problems - which
could be the cause of some of the problems you are having.


Regards

Pete


_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Einar Lygre-2

Thank you to all of you.

JavaScript was the problem and after forbidding scripts for the
problematic  websites the crashes disappeared. I have used NoScript for
"years", but of some reason I used it  most for protecting against not
so trustworthy websites, and our Norwegian newspaper could not belong to
them!
The problems with the right mouse button disappeared as well.
Now I will have problems with websites where JavaScript is necessary,
but one problem at a time.

Einar

On 12.08.13 12:49, Peter Brown wrote:

> Hi Einar
>
> Einar Lygre wrote:
>> Firefox is a great problem for me. I normally use some time to browse
>> through some Norwegian newspaper. Using one hour and having 10 crashes
>> is normal. I open 7 newspapers in 7 tabs and already during the opening
>> process FF often crashes. Restarting and the tabs are loaded. If I
>> select a link from within a website FF very often crashes. Then it may
>> help (for a time) to close other tabs and just use one. Some crashes
>> also freezes the machine, and only a very few gives any error messages.
>> Restarting FF after some of the crashes gives an apologize, other start
>> from the point of crash.
>> Now a new problem has surfaced. In two of the newspapers (www.ap.no and
>> www.bt.no) I can't use the right mouse button on a link to get up the
>> menu to open the new site in a new tab. The menu doesn't show up. These
>> two newspapers are recently bought up by a company, and I guess they now
>> are using some common software to create the sites. But this is not a
>> problem in MS Vista.
>> I should like to know if all these crashes is a common problem, or if it
>> is just how it is in eCS. I think I have done all the right things and
>> updated to latest version. New profile, new fontconfig.
>> FF 10.0.12, eCS 2.1
>>
>> Einar Lygre
>
>
>
> I'm using Seamonkey 2.7.2-12 which is, I believe, based on the Firefox
> 10.0.12 code.
>
> I cannot comment on the tabs problem as I rarely have more than 5 tabs
> open except to say that the few times I have had more than 5 tabs open
> Seamonkey has not crashed as a result. I suspect available RAM and
> system configuration may have a bearing on the problem.
>
> No problems using the 2 websites you mention; RMB on a link and the
> options menu appears. Maybe I'm trying the wrong links...
>
> I do have a useful extension installed called NoScript
> http://noscript.net/ which may help with javascript problems - which
> could be the cause of some of the problems you are having.
>
>
> Regards
>
> Pete
>
>

_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Andreas Schnellbacher
In reply to this post by Dave Yeo-3
Dave Yeo wrote:

> Andreas Schnellbacher wrote:
>
>> IMO Mozilla entered the alpha stage (for the OS/2 version) soon
>> after Peter discontinued development. Let's hope for Dmitry's
>> version.
>
> Peter had a lot of time to maintain Mozilla though he was obviously
> getting tired of it towards the end. It's a massive code base that
> could easily keep a developer employed full time maintaining the
> OS/2 port. The other problems came with the direction that Mozilla
> took after 3.7 became 4. Rapid release schedule means more changes
> and more things to break. Moving to xul means having a DLL that is
> right at the limits of our old operating system (or perhaps the
> linkers). For a long time we haven't been able to build a debug
> version or even a version with debug symbols enabled. Working with
> Steven we finally figured out that the problem was happening when
> the object size of xul.dll was over 32MBs. Hopefully Dimitri will be
> able to build debug versions of 17esr as debugging is much easier
> and also hoping that the IPC stuff doesn't push the DLL over the
> size limit.
>
> Without the IPC stuff versions 5 to 10 were sorta mickey moused
> together and became harder and harder to build, especially for
> non-programmers such Walter and me.
>
> Dimitri seems pretty knowledgeable and has learned a lot about the
> Mozilla code since he started working on it but as I said it is a
> huge complex code base that really could use a few developers
> working full time on it. It also really needs a more modern
> operating system then OS/2, we're missing the 3d stuff, some font
> stuff, large continuous chunks of memory and probably more.

Dave, of course I don't want to discredit you. To be more precise, the
problems started already in Peter's area: printing. It depends on how
fast the Mozilla code base changes. Nowadays it can't be done by a
free-time programmer. Dmitry's commits make hope and it makes fun to
read his carefully worded comments. Also Rich contributed a lot of
code to make printing work in a prior version. (I was astonished about
that when I scanned it today.)

Andreas Schnellbacher
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Dave Yeo-3
In reply to this post by Cameron Kaiser-2
Cameron Kaiser wrote:
> FWIW, these are the same problems we deal with in the TenFourFox port.
> We don't implement IPC either (OS X 10.4 doesn't have spawn() and
> friends), we don't have 3D (OpenGL 1.1 only), we don't support jemalloc
> (lots of weird bugs when we try to turn it on) and we eventually had to
> hack the OS X PowerPC linker to get libxul to build properly. For
> Firefox 19 we moved entirely to a new, non-Apple gcc toolchain with our
> hacked linker, and I do most of my work in an upgraded gdb I hacked
> myself. But the browser still gets up and runs.

We do have spawn() and friends, just that IPC is a large amount of work.
The current developer has put in quite a bit of work and FF still isn't
running. He has now gone on a tangent to replace the build system :( How
do you work around IPC on PPC? Looking at your patches I can't see much
to disable it.
We've never had hardware support for 3d and only have OpenGL 1.1, we now
use the OpenWatcom linker (hacked for native debug code)  for which
source is available but our native object format is ancient and has 16
bit parts that seem to be overflowing with XUL. Not sure what the limits
of the kernel DLL loader are either.

>
> The point being, Firefox doesn't actually *need* these things to
> function, though they do work better with them. We're fortunate in that
> Harfbuzz works pretty well on 10.4, so fonts haven't been a big problem
> for us other than altering gfx/thebes/ to use the old ATSUI font API
> instead of the current CoreText one, so we've been able to hold those
> things together and instead put effort into performance, such as a
> PowerPC JavaScript JIT and AltiVec acceleration. I don't know how well
> the x86 JIT works on OS/2, but I imagine it works at some level.

We have a fontconfig implementation that works good enough and the x86
JIT works very well as long as memory holds out. 32 bit OS/2 originally
only had a 512MB process space which after loading various DLLs means
less then 300MBs. We do have access to higher memory but it is somewhat
mickeymouse so memory allocation can be a problem.

>
> For OS/2, other than fonts, I think your biggest challenge will be the
> build system, and not just because of the present work Gregor is doing.
> If someone is able to get Python 2.7.3 to be happy on OS/2, that will go
> a long way to improving the longevity of the OS/2 port.

I'm using Python 2.7.2. Need to port virtualenv though.

> Additionally, I
> think that whoever maintains Firefox OS/2 also needs to be ruthless
> about removing features that can't be made to work if they are not
> essential. We've jettisoned plugins entirely in TenFourFox because
> they're just not workable, for example. I see you've made that choice
> for downloadable fonts, which is unfortunate, but may be a better option
> than fruitless struggles in the debugger.

The downloadable fonts should be able to be made to work, our big
problem is a shortage of knowledgeable programmers. For a while I was
the only one left who could even build it and while I've learned much
about porting and programming, I'm not a programmer.

>
> Mozilla is starting to talk more about resurrecting Electrolysis (e10s),
> for which IPC is a prerequisite. For Fx25 this is just thumbnails, so it
> can be turned off (not essential). However, the talk of better
> sandboxing presupposes detached processes, so I've started to look again
> at implementing some of the spawn() family of functions in userspace.
> OS/2 will probably need to do the same.

Most of what is needed to implement IPC is available with some of the
wide char functions missing, it's just a matter of someone spending a
month or so doing it, meanwhile keeping other stuff working, widget
code, the JIT and such. Meanwhile more projects drop support so things
like sqlite3 no longer works out of the box.
Dave
ps Interesting that you have a floodgap email address, I had to fire up
the system Gopher client which I have set to start with floodgap and I
see you maintain gopher.floodgap.com. Nice to still have some gopher
space left.
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Steve Wendt
On 8/20/2013 10:25 PM, Dave Yeo wrote:

> ps Interesting that you have a floodgap email address, I had to fire up
> the system Gopher client which I have set to start with floodgap and I
> see you maintain gopher.floodgap.com. Nice to still have some gopher
> space left.

Out of curiosity, what do you still use gopher for?

_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Cameron Kaiser-2
In reply to this post by Dave Yeo-3
On 8/20/13 10:25 PM, Dave Yeo wrote:

>> FWIW, these are the same problems we deal with in the TenFourFox port.
>> We don't implement IPC either (OS X 10.4 doesn't have spawn() and
>> friends), we don't have 3D (OpenGL 1.1 only), we don't support jemalloc
>> (lots of weird bugs when we try to turn it on) and we eventually had to
>> hack the OS X PowerPC linker to get libxul to build properly. For
>> Firefox 19 we moved entirely to a new, non-Apple gcc toolchain with our
>> hacked linker, and I do most of my work in an upgraded gdb I hacked
>> myself. But the browser still gets up and runs.
>
> We do have spawn() and friends, just that IPC is a large amount of work.
> The current developer has put in quite a bit of work and FF still isn't
> running. He has now gone on a tangent to replace the build system :(

That doesn't sound like a very productive approach :( why isn't he
simply just trying to get it to work? Dragging that along is really
going to hurt maintainability.

> How
> do you work around IPC on PPC? Looking at your patches I can't see much
> to disable it.

10.4 has just enough support for IPC *threads* that we can get that part
to fly. As far as chromium IPC in general, I've just implemented a lot
of hacks to get the parts that build to work and fortunately Mozilla
hasn't required us to implement any of the parts that don't. It's the
future multi-*process* stuff that's going to be challenging. For now,
thumbnails will be disabled on sight. Electrolysis will definitely cause
me much heartburn.

[...]

>> Additionally, I
>> think that whoever maintains Firefox OS/2 also needs to be ruthless
>> about removing features that can't be made to work if they are not
>> essential. We've jettisoned plugins entirely in TenFourFox because
>> they're just not workable, for example. I see you've made that choice
>> for downloadable fonts, which is unfortunate, but may be a better option
>> than fruitless struggles in the debugger.
>
> The downloadable fonts should be able to be made to work, our big
> problem is a shortage of knowledgeable programmers. For a while I was
> the only one left who could even build it and while I've learned much
> about porting and programming, I'm not a programmer.

Yes, and aging hardware. The guy who wrote a goodly portion of our
JaegerMonkey PowerPC implementation alas is out of commission because
his PowerBook G4 died. Right now, BaselineCompiler PowerPC is a solo effort.

>> Mozilla is starting to talk more about resurrecting Electrolysis (e10s),
>> for which IPC is a prerequisite. For Fx25 this is just thumbnails, so it
>> can be turned off (not essential). However, the talk of better
>> sandboxing presupposes detached processes, so I've started to look again
>> at implementing some of the spawn() family of functions in userspace.
>> OS/2 will probably need to do the same.
>
> Most of what is needed to implement IPC is available with some of the
> wide char functions missing, it's just a matter of someone spending a
> month or so doing it, meanwhile keeping other stuff working, widget
> code, the JIT and such. Meanwhile more projects drop support so things
> like sqlite3 no longer works out of the box.

The other option is just to comment out the parts in IPC that aren't
actually used. Mozilla imported a lot of code, but relatively few parts
of it are called, especially the abstraction layer stuff which is
already more usefully implemented in NSPR.

> ps Interesting that you have a floodgap email address, I had to fire up
> the system Gopher client which I have set to start with floodgap and I
> see you maintain gopher.floodgap.com. Nice to still have some gopher
> space left.

The main Floodgap server sits about 10 feet away from me in my house.
It's an IBM POWER6. Makes Veronica searches a breeze. :)

OverbiteFF should work just fine with OS/2 Firefox since it's 100%
JavaScript, no binary modules.

Cameron Kaiser

_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Cameron Kaiser-2
In reply to this post by Steve Wendt
On 8/21/13 10:36 AM, Steve Wendt wrote:
> On 8/20/2013 10:25 PM, Dave Yeo wrote:
>
>> ps Interesting that you have a floodgap email address, I had to fire up
>> the system Gopher client which I have set to start with floodgap and I
>> see you maintain gopher.floodgap.com. Nice to still have some gopher
>> space left.
>
> Out of curiosity, what do you still use gopher for?
>
I don't know if this was to Dave or me, but I use gopher as a minimal
communication protocol. For example, my home automation systems and the
server room climate control system report their data over an internal
Gopher server. I also use it as an easy one-way file server for my older
systems, and I've put a lot of those old file archives online for others
to browse as well.

Also, thanks to NOAA data, weather reports over Gopher are so much
faster and more convenient than Weather Underground or the Weather
Channel. :)

Cameron Kaiser
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Dave Yeo-3
In reply to this post by Steve Wendt
Steve Wendt wrote:
> On 8/20/2013 10:25 PM, Dave Yeo wrote:
>
>> ps Interesting that you have a floodgap email address, I had to fire up
>> the system Gopher client which I have set to start with floodgap and I
>> see you maintain gopher.floodgap.com. Nice to still have some gopher
>> space left.
>
> Out of curiosity, what do you still use gopher for?
>

Mostly nostalgia, there is still some interesting stuff out there and it
is much more low-bandwidth friendly.
Dave
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Dave Yeo-3
In reply to this post by Cameron Kaiser-2
Cameron Kaiser wrote:

> On 8/20/13 10:25 PM, Dave Yeo wrote:
>>> FWIW, these are the same problems we deal with in the TenFourFox port.
>>> We don't implement IPC either (OS X 10.4 doesn't have spawn() and
>>> friends), we don't have 3D (OpenGL 1.1 only), we don't support jemalloc
>>> (lots of weird bugs when we try to turn it on) and we eventually had to
>>> hack the OS X PowerPC linker to get libxul to build properly. For
>>> Firefox 19 we moved entirely to a new, non-Apple gcc toolchain with our
>>> hacked linker, and I do most of my work in an upgraded gdb I hacked
>>> myself. But the browser still gets up and runs.
>>
>> We do have spawn() and friends, just that IPC is a large amount of work.
>> The current developer has put in quite a bit of work and FF still isn't
>> running. He has now gone on a tangent to replace the build system :(
>
> That doesn't sound like a very productive approach :( why isn't he
> simply just trying to get it to work? Dragging that along is really
> going to hurt maintainability.

Some people hate the auto tools and he seems to think the build system
is responsible for not being able to build a working debug build (it
actually seems to be a system limitation, XUL is huge)
I agree about the maintainability.

>
>> How
>> do you work around IPC on PPC? Looking at your patches I can't see much
>> to disable it.
>
> 10.4 has just enough support for IPC *threads* that we can get that part
> to fly. As far as chromium IPC in general, I've just implemented a lot
> of hacks to get the parts that build to work and fortunately Mozilla
> hasn't required us to implement any of the parts that don't.

I'll have to retry. Previously the show stopper was no working vswprintf().

> It's the
> future multi-*process* stuff that's going to be challenging. For now,
> thumbnails will be disabled on sight. Electrolysis will definitely cause
> me much heartburn.

Yes, even with working IPC it looks like quite a bit of system dependent
code will be required.

>
> [...]
>>> Additionally, I
>>> think that whoever maintains Firefox OS/2 also needs to be ruthless
>>> about removing features that can't be made to work if they are not
>>> essential. We've jettisoned plugins entirely in TenFourFox because
>>> they're just not workable, for example. I see you've made that choice
>>> for downloadable fonts, which is unfortunate, but may be a better option
>>> than fruitless struggles in the debugger.
>>
>> The downloadable fonts should be able to be made to work, our big
>> problem is a shortage of knowledgeable programmers. For a while I was
>> the only one left who could even build it and while I've learned much
>> about porting and programming, I'm not a programmer.
>
> Yes, and aging hardware. The guy who wrote a goodly portion of our
> JaegerMonkey PowerPC implementation alas is out of commission because
> his PowerBook G4 died. Right now, BaselineCompiler PowerPC is a solo
> effort.

Well OS/2 (actually eComstation, an OEM release) will run on some modern
hardware though it runs better in a virtual machine in most cases as
things don't have to be set to legacy mode.
My computer died the other week, just moved the hard drives to a P4
spare that was handy and had to change the Ethernet driver. OS/2 is very
good about moving it from machine to machine, I'm typing this on an
install I did in '97 or so :)
Of course the P4 feels slow and only has 768 MBs of memory. I could
build Firefox in 50 minutes on the core2duo using both cores. This box
takes 4 hours :(

>
[...]
>
> The other option is just to comment out the parts in IPC that aren't
> actually used. Mozilla imported a lot of code, but relatively few parts
> of it are called, especially the abstraction layer stuff which is
> already more usefully implemented in NSPR.

I'll try this as Dimitri has written a vswprintf() implementation.

>
>> ps Interesting that you have a floodgap email address, I had to fire up
>> the system Gopher client which I have set to start with floodgap and I
>> see you maintain gopher.floodgap.com. Nice to still have some gopher
>> space left.
>
> The main Floodgap server sits about 10 feet away from me in my house.
> It's an IBM POWER6. Makes Veronica searches a breeze. :)
>
> OverbiteFF should work just fine with OS/2 Firefox since it's 100%
> JavaScript, no binary modules.
>

The native Gopher client works quite well with everything I've tried
though I had to add a png viewer :). Even supports CSO/ph, not sure
about hURLs as I couldn't find a good description but for the example of
adding a ssh client, it is trivial to add helpers though perhaps not in
URL ormat
Dave
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Dave Yeo-3
In reply to this post by Andreas Schnellbacher
Andreas Schnellbacher wrote:
> To be more precise, the
> problems started already in Peter's area: printing. It depends on how
> fast the Mozilla code base changes. Nowadays it can't be done by a
> free-time programmer. Dmitry's commits make hope and it makes fun to
> read his carefully worded comments. Also Rich contributed a lot of
> code to make printing work in a prior version. (I was astonished about
> that when I scanned it today.)

The problems started with the switch to the Cairo rendering engine.
Doodle got that working in a good enough manner but as you note,
printing no longer worked. Rich did a lot of fixes (almost a rewrite) to
Cairo including adding an OS2 print surface that resurrected printing
and sped up general PM drawing. He did get paid to fix the printing,
other stuff such as the excellent drag'n'drop support he did on his
time. Unluckily he also seems to have moved on.
Dave
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2
Reply | Threaded
Open this post in threaded view
|

Re: Firefox's strange behavior

Dave Yeo-3
In reply to this post by Einar Lygre-2
Einar Lygre wrote:

> Thank you to all of you.
>
> JavaScript was the problem and after forbidding scripts for the
> problematic  websites the crashes disappeared. I have used NoScript for
> "years", but of some reason I used it  most for protecting against not
> so trustworthy websites, and our Norwegian newspaper could not belong to
> them!
> The problems with the right mouse button disappeared as well.
> Now I will have problems with websites where JavaScript is necessary,
> but one problem at a time.

A lot of the time it is JavaScript from a third party such as
advertisers or of course Flash (I ended up moving the Flash plugin out
of the way here) that causes the problems so sometimes you can get away
with just enabling some scripts. Of course well written sites should
work satisfactory without JavaScript but...
Dave
_______________________________________________
dev-ports-os2 mailing list
[hidden email]
https://lists.mozilla.org/listinfo/dev-ports-os2