10:47 a.m.
Hello!
Recently I've succeeded in cross-compiling Netsurf GTK for my iPaq h2200
PDA with Angstrom linux ( http://www.angstrom-distribution.org )
using OpenEmbedded ( http://www.openembedded.org/ ).
It works ok, but there's a little problem - scrollbars doesn't fit
in tiny QVGA (240x320) screen, making netsurf nearly unusable. Here's
screenshot:
http://mad.is-a-geek.org/netsurf/netsurf.png
On "big PC" netsurf's window can't be resized less than about 488
pixels
width, so I guessed that it's hardcoded somewhere. After some poking at
the code i finally found the place - it was in glade file, the width of
statusbar label, measured in chars! After removing this property of label,
scrollbars finally appeared:
http://mad.is-a-geek.org/netsurf/netsurf-patched.png
Here's my patch:
Index: gtk/res/netsurf.glade
===================================================================
--- gtk/res/netsurf.glade (revision 3862)
+++ gtk/res/netsurf.glade (working copy)
@@ -799,7 +799,6 @@
<property name="label"
translatable="yes">Status bar text goes here</property>
<property name="selectable">True</property>
<property
name="ellipsize">PANGO_ELLIPSIZE_MIDDLE</property>
- <property name="width_chars">50</property>
</widget>
<packing>
<property name="resize">False</property>
I'm planning to submit my building recipes to OpenEmbedded soon,
so other PDA users could enjoy Netsurf on their PDAs too.
--
Best regards
Dmitry "MAD" Artamonow
P.S.
You can see some other screenshots of Netsurf on h2200 on my page here:
http://mad.is-a-geek.org/netsurf/
11:08 a.m.
On Sat, 2008-02-23 at 18:47 +0300, Dmitry Artamonow wrote:
<snip>
On "big PC" netsurf's window can't be resized less
than about 488 pixels
width, so I guessed that it's hardcoded somewhere. After some poking at
the code i finally found the place - it was in glade file, the width of
statusbar label, measured in chars! After removing this property of label,
scrollbars finally appeared:
http://mad.is-a-geek.org/netsurf/netsurf-patched.png
Looks good: I'm sure there's a reason that 50 is there (I seem to recall
some versions of GTK making that label invisibly tiny by default mean
people don't realise it's even there!).
I'm currently at FOSDEM, and for some reason I can't get to NetSurf's
Subversion from ULB, the host university. I'll try to investigate this
when I get back to the UK or sooner if I can, and apply your patch to
our repository.
I'm planning to submit my building recipes to OpenEmbedded soon,
so other PDA users could enjoy Netsurf on their PDAs too.
Yes: the reason I originally got involved in the project was to have a
faster browser for my NetBook Pro.
B.
3:52 p.m.
On 16:08 Sat 23 Feb , Rob Kendrick wrote:
On Sat, 2008-02-23 at 18:47 +0300, Dmitry Artamonow wrote:
<snip>
> On "big PC" netsurf's window can't be resized less than about 488
pixels
> width, so I guessed that it's hardcoded somewhere. After some poking at
> the code i finally found the place - it was in glade file, the width of
> statusbar label, measured in chars! After removing this property of label,
> scrollbars finally appeared:
> http://mad.is-a-geek.org/netsurf/netsurf-patched.png
Looks good: I'm sure there's a reason that 50 is there (I seem to recall
some versions of GTK making that label invisibly tiny by default mean
people don't realise it's even there!).
Well, changing 50 to 20 works too
for 240x320 screen (QVGA portrait),
but i guess it depends on fonts/theme/something else.
I don't have old gtk to test - but on both my PC (gtk-2.12.1) and PDA
(gtk-2.10.14) there's no problem with label visibility when removing
width_chars completely.
I'm currently at FOSDEM, and for some reason I can't get to
NetSurf's
Subversion from ULB, the host university. I'll try to investigate this
when I get back to the UK or sooner if I can, and apply your patch to
our repository.
> I'm planning to submit my building recipes to OpenEmbedded soon,
> so other PDA users could enjoy Netsurf on their PDAs too.
Yes: the reason I originally got involved in the project was to have a
faster browser for my NetBook Pro.
FYI. I've submitted my recipe to OE bugzilla here:
http://bugs.openembedded.net/show_bug.cgi?id=3857
And sent message to angstrom-devel list:
http://lists.linuxtogo.org/pipermail/angstrom-distro-devel/2008-February/...
B.
12:07 p.m.
I built it manually for my Zaurus SL-C3100 runnning Angstrom, but found
it really slow. Do you have the same issue?
All the same, thank for making a bb file! Hopefully it will trickle
into the stable branch relatively soon.
On Sat, 23 Feb 2008 18:47:37 +0300
Dmitry Artamonow <mad_soft(a)inbox.ru> wrote:
Hello!
Recently I've succeeded in cross-compiling Netsurf GTK for my iPaq
h2200 PDA with Angstrom linux
( http://www.angstrom-distribution.org ) using OpenEmbedded
( http://www.openembedded.org/ ).
It works ok, but there's a little problem - scrollbars doesn't fit
in tiny QVGA (240x320) screen, making netsurf nearly unusable. Here's
screenshot:
http://mad.is-a-geek.org/netsurf/netsurf.png
On "big PC" netsurf's window can't be resized less than about 488
pixels width, so I guessed that it's hardcoded somewhere. After some
poking at the code i finally found the place - it was in glade file,
the width of statusbar label, measured in chars! After removing this
property of label, scrollbars finally appeared:
http://mad.is-a-geek.org/netsurf/netsurf-patched.png
Here's my patch:
Index: gtk/res/netsurf.glade
===================================================================
--- gtk/res/netsurf.glade (revision 3862)
+++ gtk/res/netsurf.glade (working copy)
@@ -799,7 +799,6 @@
<property name="label"
translatable="yes">Status
bar text goes here</property> <property
name="selectable">True</property> <property
name="ellipsize">PANGO_ELLIPSIZE_MIDDLE</property>
- <property name="width_chars">50</property>
</widget>
<packing>
<property name="resize">False</property>
I'm planning to submit my building recipes to OpenEmbedded soon,
so other PDA users could enjoy Netsurf on their PDAs too.
--
Best regards
Dmitry "MAD" Artamonow
P.S.
You can see some other screenshots of Netsurf on h2200 on my page
here: http://mad.is-a-geek.org/netsurf/
11:11 a.m.
On Sat, 2008-02-23 at 11:07 -0600, Daniel Taylor wrote:
I built it manually for my Zaurus SL-C3100 runnning Angstrom, but
found
it really slow. Do you have the same issue?
How is it slow? Slow to fetch, slow to render, slow to scroll, etc?
Did you compile NetSurf without -g and with -O2? It'll be very slow on
ARM without them.
B.
1:19 p.m.
I'll give it another shot. I usually use -O3, but -g may have been in
there. Some other folks have pestered me for more NetSUrf comparisons,
so I guess I'll just have to do it.
I'll keep you informed.
On Sat, 23 Feb 2008 16:11:01 +0000
Rob Kendrick <rjek(a)netsurf-browser.org> wrote:
On Sat, 2008-02-23 at 11:07 -0600, Daniel Taylor wrote:
> I built it manually for my Zaurus SL-C3100 runnning Angstrom, but
> found it really slow. Do you have the same issue?
How is it slow? Slow to fetch, slow to render, slow to scroll, etc?
Did you compile NetSurf without -g and with -O2? It'll be very slow
on ARM without them.
B.
5:09 p.m.
On Sat, 2008-02-23 at 12:19 -0600, Daniel Taylor wrote:
I'll give it another shot. I usually use -O3, but -g may have
been in
there. Some other folks have pestered me for more NetSUrf comparisons,
so I guess I'll just have to do it.
Don't use -O3 on ARMs with puny caches: in the long run it'll be a loss.
B.
6:40 p.m.
I'm getting errors building it now for some reason:
parser.c- undefined reference to 'assert'
objects-gtk/mng.c- undefined reference to <lots of mng stuff>
Any ideas? I have libmng installed (built it from source, all of the
dev stuff is there).
On Sat, 23 Feb 2008 22:09:09 +0000
Rob Kendrick <rjek(a)netsurf-browser.org> wrote:
On Sat, 2008-02-23 at 12:19 -0600, Daniel Taylor wrote:
> I'll give it another shot. I usually use -O3, but -g may have been
> in there. Some other folks have pestered me for more NetSUrf
> comparisons, so I guess I'll just have to do it.
Don't use -O3 on ARMs with puny caches: in the long run it'll be a
loss.
B.
5:46 p.m.
On Sat, 2008-02-23 at 17:40 -0600, Daniel Taylor wrote:
I'm getting errors building it now for some reason:
parser.c- undefined reference to 'assert'
objects-gtk/mng.c- undefined reference to <lots of mng stuff>
Any ideas? I have libmng installed (built it from source, all of the
dev stuff is there).
What have you changed since you successfully managed to build NetSurf?
B.
12:01 p.m.
I don't believe I have changed anything. I assume there have been minor
upgrades in miscellaneous libraries and so forth. I also edited the
makefile so it has no -g option and "-O" was changed to "-O2"
On Sat, 23 Feb 2008 22:46:30 +0000
Rob Kendrick <rjek(a)netsurf-browser.org> wrote:
On Sat, 2008-02-23 at 17:40 -0600, Daniel Taylor wrote:
> I'm getting errors building it now for some reason:
>
> parser.c- undefined reference to 'assert'
> objects-gtk/mng.c- undefined reference to <lots of mng stuff>
>
> Any ideas? I have libmng installed (built it from source, all of the
> dev stuff is there).
What have you changed since you successfully managed to build NetSurf?
B.
8:57 p.m.
I got it built, but it still feels slow, taking like 30 seconds to load
a fairly simple page.
It was worth a shot, though.
On Sun, 24 Feb 2008 23:17:22 +0000
Rob Kendrick <rjek(a)netsurf-browser.org> wrote:
On Sun, 2008-02-24 at 11:01 -0600, Daniel Taylor wrote:
> I don't believe I have changed anything.
Then your computer is non-deterministic, and not very useful :)
B.
8:03 p.m.
On Sun, 24 Feb 2008 19:57:53 -0600, Daniel Taylor <dannytaylor(a)visi.com>
wrote:
I got it built, but it still feels slow, taking like 30 seconds to
load
a fairly simple page.
Can you provide us with more details (such as the URL of the page in
question)? There's no way to solve the problems you are seeing otherwise.
John.
11:46 p.m.
A couple of the pages I tried:
http://www.oesf.org/forum/ (relatively slow)
http://www.linux.com/ (very slow)
http://www.tyrannozaurus.com/slashdot (decent, but specially formatted
for PDA vieweing)
I have the cache set to 100 MB, disabled AA, and have connections set to
4.
The system in quesrion is a 416 MHz ARMV5TE with 64MB RAM, plenty of
swap, and rendering at 640x480x16.
On Mon, 25 Feb 2008 01:03:59 +0000
John-Mark Bell <jmb(a)netsurf-browser.org> wrote:
On Sun, 24 Feb 2008 19:57:53 -0600, Daniel Taylor
<dannytaylor(a)visi.com> wrote:
> I got it built, but it still feels slow, taking like 30 seconds to
> load a fairly simple page.
Can you provide us with more details (such as the URL of the page in
question)? There's no way to solve the problems you are seeing
otherwise.
John.
4:23 p.m.
In article <20080224224616.2f12e93e(a)Fish.domain.actdsltmp>,
Daniel Taylor <dannytaylor(a)visi.com> wrote:
A couple of the pages I tried:
http://www.oesf.org/forum/ (relatively slow)
http://www.linux.com/ (very slow)
http://www.tyrannozaurus.com/slashdot (decent, but specially formatted
for PDA vieweing)
I have the cache set to 100 MB
There's no need to make the cache that big, really.
The system in quesrion is a 416 MHz ARMV5TE with 64MB RAM, plenty of
swap, and rendering at 640x480x16.
Well, it does seem odd that it's so slow. Here's some timings from my
machines:
Machine 1:
600 MHz XScale 80321, with 512 MB RAM. 1680x1050 32bpp screen
http://www.oesf.org/forum/ 4.1s (As reported by NetSurf's status bar)
http://www.linux.com/ 5.8s (As reported by NetSurf's status bar)
Choices: Maximum fetchers: 24
Fetches per host: 5
Cached connections: 6
Cache size: 6.4MB
Machine 2:
202 MHz StrongARM, with 98 MB old slow RAM. 1024x768 32K colour screen
(A 14 year old computer upgraded with a 12 year old CPU)
http://www.oesf.org/forum/ 6.9s (As reported by NetSurf's status bar)
http://www.linux.com/ 12.1s (As reported by NetSurf's status bar)
Choices: Maximum fetchers: 24
Fetches per host: 4
Cached connections: 6
Cache size: 2MB
All times are from starting the fetch to end of final display with all
images, etc.
I'd expect roughly something between those timings on your hardware,
probably nearer Machine 1's speed.
Best regards,
--
Michael Drake (tlsa) http://www.netsurf-browser.org/
7:27 a.m.
I'm getting 12.6 seconds (as reported by the status bar) and 32
seconds (as above) for OESF and Linux.com, respectively.
What options did you use for building it?
On Tue, 26 Feb 2008 21:23:10 +0000 (GMT)
Michael Drake <tlsa(a)netsurf-browser.org> wrote:
In article <20080224224616.2f12e93e(a)Fish.domain.actdsltmp>,
Daniel Taylor <dannytaylor(a)visi.com> wrote:
> A couple of the pages I tried:
> http://www.oesf.org/forum/ (relatively slow)
> http://www.linux.com/ (very slow)
> http://www.tyrannozaurus.com/slashdot (decent, but specially
> formatted for PDA vieweing)
> I have the cache set to 100 MB
There's no need to make the cache that big, really.
> The system in quesrion is a 416 MHz ARMV5TE with 64MB RAM, plenty of
> swap, and rendering at 640x480x16.
Well, it does seem odd that it's so slow. Here's some timings from my
machines:
Machine 1:
600 MHz XScale 80321, with 512 MB RAM. 1680x1050 32bpp screen
http://www.oesf.org/forum/ 4.1s (As reported by NetSurf's status
bar) http://www.linux.com/ 5.8s (As reported by NetSurf's
status bar)
Choices: Maximum fetchers: 24
Fetches per host: 5
Cached connections: 6
Cache size: 6.4MB
Machine 2:
202 MHz StrongARM, with 98 MB old slow RAM. 1024x768 32K colour
screen (A 14 year old computer upgraded with a 12 year old CPU)
http://www.oesf.org/forum/ 6.9s (As reported by NetSurf's status
bar) http://www.linux.com/ 12.1s (As reported by NetSurf's
status bar)
Choices: Maximum fetchers: 24
Fetches per host: 4
Cached connections: 6
Cache size: 2MB
All times are from starting the fetch to end of final display with all
images, etc.
I'd expect roughly something between those timings on your hardware,
probably nearer Machine 1's speed.
Best regards,
12:54 p.m.
On Thu, 2008-02-28 at 06:27 -0600, Daniel Taylor wrote:
I'm getting 12.6 seconds (as reported by the status bar) and 32
seconds (as above) for OESF and Linux.com, respectively.
What options did you use for building it?
The autobuilder for producing RISC OS binaries uses -O (not -O2 or
anything, I'm not entirely sure why this is) and the compiler tunes for
XScale by default.
(Also, please try to avoid top-quoting here.)
B.
11:16 a.m.
OK, I will try adding march, mtune, and mcpu lines and see what that
does.
What is "top-quoting?" I'll do my best to avoid it when I know what it
is.
4:36 p.m.
In article <20080228101640.12594f79(a)Fish.at.fishnet>,
Daniel Taylor <dannytaylor(a)visi.com> wrote:
What is "top-quoting?" I'll do my best to avoid it when
I know what it
is.
It's normally called "top-posting".
It means putting your reply above the text you're quoting. The normal
style here is to write your reply below the bit of the post you're
actually responding to and removing any bits of the original mail that
you're not responding to.
See "inline replying" here:
http://en.wikipedia.org/wiki/Top-posting
Best regards,
--
Michael Drake (tlsa) http://www.netsurf-browser.org/
4:45 p.m.
On Thu, 2008-02-28 at 10:37 -0600, Daniel Taylor wrote:
Gotcha. Using march, mtune, and mcpu doesn't help, for what
it's worth.
Is it possible to run NetSurf from a terminal on your device? If so,
can you run it by prefixing "time" onto the beginning of the command?
ie:
time ./netsurf
Then, fetch a page that seems to be taking a while, then quit NetSurf.
Your shell should then have the amount of wall clock, user and system
time NetSurf consumed - please post that here along with the URL to the
site you visited.
B.
12:35 p.m.
On Thu, 28 Feb 2008 21:45:52 +0000
Rob Kendrick <rjek(a)netsurf-browser.org> wrote:
Is it possible to run NetSurf from a terminal on your device? If so,
can you run it by prefixing "time" onto the beginning of the command?
ie:
time ./netsurf
Then, fetch a page that seems to be taking a while, then quit NetSurf.
Your shell should then have the amount of wall clock, user and system
time NetSurf consumed - please post that here along with the URL to
the site you visited.
B.
Sure.
With the mtune and stuff, going to www.linux.com:
real 1m 22.99s
user 1m 20.22s
sys 0m 0.64s
7:48 p.m.
On Thu, 2008-02-28 at 11:35 -0600, Daniel Taylor wrote:
> Is it possible to run NetSurf from a terminal on your device?
If so,
> can you run it by prefixing "time" onto the beginning of the command?
> ie:
>
> time ./netsurf
With the mtune and stuff, going to www.linux.com:
real 1m 22.99s
user 1m 20.22s
sys 0m 0.64s
Right, it really is NetSurf eating CPU time, rather than some other
bottleneck such as getting data from the kernel. Without having such a
device to hand, I've run out of suggestions for what it might be. I
need to get my hands on such a device and do some profiling and
monitoring to find out what's going on.
B.
7:39 a.m.
In article <20080228113525.7112407b(a)Fish.at.fishnet>,
With the mtune and stuff, going to www.linux.com:
real 1m 22.99s
user 1m 20.22s
sys 0m 0.64s
Are you using NetSurf 1.1 or the latest version in SVN (currently r3874)?
--
Michael Drake (tlsa) http://www.netsurf-browser.org/
5:09 a.m.
On Sat, 2008-03-01 at 13:47 -0600, Daniel Taylor wrote:
On Sat, 01 Mar 2008 12:39:55 +0000 (GMT)
>
> Are you using NetSurf 1.1 or the latest version in SVN (currently
> r3874)?
>
1.1
Please try again with the current Subversion HEAD. It has received many
improvements since the release of 1.1.
B.
10:12 a.m.
On Sun, 02 Mar 2008 10:09:21 +0000
Rob Kendrick <rjek(a)netsurf-browser.org> wrote:
Please try again with the current Subversion HEAD. It has received
many improvements since the release of 1.1.
It feels a bit faster (maybe 5 seconds or so shaved off loading the
previously mentioned pages). Still not quite usable, though.
It seems to render the page, then replace it with a blank screen a
couple of times before finally rendering the page. Why is that?
Ex:
1. load
2. flash rendered page
3. blank page and load more
4. Repeat steps 2 & 3 once or twice
5. Finally have rendered page
4953
days inactive
4961
days old
netsurf-dev@netsurf-browser.org
28 comments
5 participants
participants (5)
-
Daniel Taylor -
Dmitry Artamonow -
John-Mark Bell -
Michael Drake -
Rob Kendrick