AROS World Exec
Development => Development (General) => Topic started by: ch1ll on October 24, 2018, 02:07:53 PM
-
Hi all,
I modified a bit the source code of Odyssey from deadwood to compile with gcc-4.6.4 or gcc-6.1.0 and binutils 2.25 in the current ABI_V0. Now I can compile it cleanly, but I still cannot link it.
I see that the linker take a lot of time and use all the memory available and at the end I guess it got interrupted with the error below.
I tried with different VM running ubuntu 16.04 and 14.04 and 8 GB or RAM and with my host running Fedora 28 with 16 GB of RAM + 20 GB of swap. I always have this error. I guess that available RAM is not the real issue.
You can find my modifications on: https://github.com/nscendoni/OdysseyWebBrowser/tree/odyssey-r187682.
[100%] Building CXX object Tools/OdysseyWebBrowser/CMakeFiles/owb.dir/main.cpp.obj
cd /home/vagrant/AROS-20180827-source/bin/linux-i386/gen/local/odyssey/Tools/OdysseyWebBrowser && /home/vagrant/AROS-20180827-source/bin/linux-x86_64/tools/crosstools/i386-aros-g++ -DBUILDING_WITH_CMAKE=1 -DHAVE_CONFIG_H=1 -DUSER_AGENT_GTK_MAJOR_VERSION=601 -DUSER_AGENT_GTK_MINOR_VERSION=1 --sysroot /home/vagrant/AROS-20180827-source/bin/linux-i386/AROS/Development --sysroot /home/vagrant/AROS-20180827-source/bin/linux-i386/AROS/Development -I/home/vagrant/AROS-20180827-source/bin/linux-i386/AROS/Development/include -I -Wall -DAROS_BUILD_TYPE=AROS_BUILD_TYPE_PERSONAL -g -DADEBUG=1 -DMDEBUG=1 -O0 -m32 -march=i686 -Wno-unused-function -Wno-conversion-null -Wno-deprecated-declarations -Wno-cast-align -std=c++11 -O3 -DNDEBUG -fno-exceptions -fno-strict-aliasing -fno-rtti -O2 -I/home/vagrant/AROS-20180827-source/bin/linux-i386/AROS/Development/include/cairo -I/home/vagrant/AROS-20180827-source/local/odyssey/Source/WebKit/mui/Api/MorphOS -I/home/vagrant/AROS-20180827-source/local/odyssey/Source/WebKit/mui/Api/AROS/include -I/home/vagrant/AROS-20180827-source/bin/linux-i386/gen/local/odyssey/DerivedSources/WebCore -o CMakeFiles/owb.dir/main.cpp.obj -c /home/vagrant/AROS-20180827-source/local/odyssey/Tools/OdysseyWebBrowser/main.cpp
Linking CXX executable ../../bin/owb
cd /home/vagrant/AROS-20180827-source/bin/linux-i386/gen/local/odyssey/Tools/OdysseyWebBrowser && /usr/bin/cmake -E cmake_link_script CMakeFiles/owb.dir/link.txt --verbose=1
/home/vagrant/AROS-20180827-source/bin/linux-x86_64/tools/crosstools/i386-aros-g++ --sysroot /home/vagrant/AROS-20180827-source/bin/linux-i386/AROS/Development --sysroot /home/vagrant/AROS-20180827-source/bin/linux-i386/AROS/Development -I/home/vagrant/AROS-20180827-source/bin/linux-i386/AROS/Development/include -I -Wall -DAROS_BUILD_TYPE=AROS_BUILD_TYPE_PERSONAL -g -DADEBUG=1 -DMDEBUG=1 -O0 -m32 -march=i686 -Wno-unused-function -Wno-conversion-null -Wno-deprecated-declarations -Wno-cast-align -std=c++11 -O3 -DNDEBUG -fno-exceptions -fno-strict-aliasing -fno-rtti -O2 -L -static-libstdc++ -g CMakeFiles/owb.dir/main.cpp.obj -o ../../bin/owb ../../lib/libjavascriptcoremui.a ../../lib/libWebCoreMUI.a ../../lib/libwebkitmui.a -lz -lcairo -lpixman -lfontconfig -lfreetype2 -lxml2 -lz -lcurl -ljpeg -lpng -lssl -lcrypto -licui18n -licuuc -licudata -lsqlite3 -lxslt -ldebug -lopenurl -lrexxsyslib -lavformat -lbz2 -lavcodec -liconv -lz -lavutil -lswresample -lswscale ../../lib/libWebCoreMUI.a ../../lib/libjavascriptcoremui.a /home/vagrant/AROS-20180827-source/bin/linux-i386/AROS/Development/lib/libfontconfig.a /home/vagrant/AROS-20180827-source/bin/linux-i386/AROS/Development/lib/libfreetype2.a /home/vagrant/AROS-20180827-source/bin/linux-i386/AROS/Development/lib/libz.a ../../lib/libWTFMUI.a -lcairo -lpixman -lfontconfig -lfreetype2 -lxml2 -lcurl -ljpeg -lpng -lssl -lcrypto -licui18n -licuuc -licudata -lsqlite3 -lxslt -ldebug -lopenurl -lrexxsyslib -lavformat -lbz2 -lavcodec -liconv -lavutil -lswresample -lswscale
nm: /tmp/catmpl76Y8i: File format not recognized
/home/vagrant/AROS-20180827-source/bin/linux-x86_64/tools/crosstools/i386-aros-objdump: /tmp/catmpl76Y8i: File format not recognized
/tmp/catmpl76Y8i: file not recognized: File format not recognized
-
cant really tell whats wrong. maybe the hard disk is full and the file got truncated. sorry.
-
Strange, you don't need nm or objdump for linking.
Did you take a look at ../../bin/owb?
-
Hi all
I tried again with a VM with 100GB of disk, 8 GB of RAM + 17 of swap. I checked the system during the link, and it used all the ram + 2.5 GB of swap and then it finished with th same output. I had 57 GB of free space during the linking.
I have the strace of that process. I'm reading it but unfortunately I don't know details about the internals of binutils. If somebody has some suggestion please let me know.
Cheers
Ch1ll
-
I will ask Neil. Thanks for your work and hope eventually it will work.
-
Hi all,
I have some news. I tried to compile using a VM with Ubuntu 14 32bit and it looks that I do not have that error, but ffmpeg still cannot find some symbol. I have now 2 questions:
1. My understanding is that bsdsocket.library is a core library and should be automatically linked, isn't it?
2. log2, log2f and __isinff are defined in Development/include/aros/stdc/math.h do you have any idea why they are not found?
nicola@ubuntu:~/AROS-20181022-source/bin/linux-i386/gen/local/odyssey/Tools/OdysseyWebBrowser$ /home/nicola/AROS-20181022-source/bin/linux-i386/tools/crosstools/i386-aros-g++ --sysroot /home/nicola/AROS-20180827-source/bin/linux-i386/AROS/Development --sysroot /home/nicola/AROS-20181022-source/bin/linux-i386/AROS/Development -I/home/nicola/AROS-20181022-source/bin/linux-i386/AROS/Development/include -I -Wall -DAROS_BUILD_TYPE=AROS_BUILD_TYPE_PERSONAL -O2 -m32 -march=i686 -Wno-unused-function -Wno-conversion-null -Wno-deprecated-declarations -Wno-cast-align -std=c++11 -O3 -DNDEBUG -fno-exceptions -fno-strict-aliasing -fno-rtti -O2 -L -static-libstdc++ CMakeFiles/owb.dir/main.cpp.obj -o ../../bin/owb ../../lib/libjavascriptcoremui.a ../../lib/libWebCoreMUI.a ../../lib/libwebkitmui.a -lz -lcairo -lpixman -lfontconfig -lfreetype2 -lxml2 -lz -lcurl -ljpeg -lpng -lssl -lcrypto -licui18n -licuuc -licudata -lsqlite3 -lxslt -ldebug -lopenurl -lrexxsyslib -lavformat -lbz2 -lavcodec -liconv -lz -lavutil -lswresample -lswscale ../../lib/libWebCoreMUI.a ../../lib/libjavascriptcoremui.a /home/nicola/AROS-20181022-source/bin/linux-i386/AROS/Development/lib/libfontconfig.a /home/nicola/AROS-20181022-source/bin/linux-i386/AROS/Development/lib/libfreetype2.a /home/nicola/AROS-20181022-source/bin/linux-i386/AROS/Development/lib/libz.a ../../lib/libWTFMUI.a -lcairo -lpixman -lfontconfig -lfreetype2 -lxml2 -lcurl -ljpeg -lpng -lssl -lcrypto -licui18n -licuuc -licudata -lsqlite3 -lxslt -ldebug -lopenurl -lrexxsyslib -lavformat -lbz2 -lavcodec -liconv -lavutil -lswresample -lswscale -lpthread -lz -lm
There are undefined symbols in '../../bin/owb':
U CloseSocket /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavformat/udp.c:977
U connect /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavformat/udp.c:460
U __isinff /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavcodec/nellymoserenc.c:267
U log2 /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavcodec/metasound.c:78
U log2f /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavcodec/imc.c:429
U recv /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavformat/udp.c:945
U send /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavformat/udp.c:966
U sendto /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavformat/udp.c:962
-
Hi
precisely what did you change from the original code?
have a nice Sunday
-
Hi sabbate,
you can see details here: https://github.com/deadwood-pl/OdysseyWebBrowser/compare/odyssey-r187682...nscendoni:odyssey-r187682?expand=1
Basically I commented some ASSERT and made some small modifications to compile with gcc-6.1.0
-
hi morning
I think that there is need of an update regarding the openssl library for the proper functioning of the browser, has already been ported and it is available, if it is discussed on the old forum, thanks for your contribution
thank you
salvatore
-
Hi Salvatore
yes, this is why I was compiling it! :)
-
Awesome! Can you please port the best Open Source Creativity Software Programs to AROS ABI_V0 like GIMP, Blender, Audacity, Inkscape, OpenShot, Mypaint, Cura, stopmotion, etc.
Also, don't forget to port VLC as well to AORS!
Thank you for considering this request
-
Awesome! Can you please port the best Open Source Creativity Software Programs to AROS ABI_V0 like GIMP, Blender, Audacity, Inkscape, OpenShot, Mypaint, Cura, stopmotion, etc.
Also, don't forget to port VLC as well to AORS!
Thank you for considering this request
Don't be stupid. Try to appreciate what is being done.
-
@ch1ll
thanks, in the old forum there was talk of a possible new version of ffmpeg, I do not know if it was brought and then implemented
salvatore
-
Hi all,
About ffmpeg upgrade we will see once the compilation works properly, and about other Open Source software on AROS let's discuss in other threads.
A small upgrade on the linking issue. It looks to me that now only the symbols from bsdsocket are not found on ffmpeg. I see that OpenLibrary() is never called.
I guess that this is required, and I guess this can be the reason why I cannot link the bsdsocket symbols. Is that correct?
Here the situation now:
nicola@ubuntu:~/AROS-20181022-source/bin/linux-i386/gen/local/odyssey/Tools/OdysseyWebBrowser$ /home/nicola/AROS-20181022-source/bin/linux-i386/tools/crosstools/i386-aros-g++ --sysroot /home/nicola/AROS-20180827-source/bin/linux-i386/AROS/Development --sysroot /home/nicola/AROS-20181022-source/bin/linux-i386/AROS/Development -I/home/nicola/AROS-20181022-source/bin/linux-i386/AROS/Development/include -I -Wall -DAROS_BUILD_TYPE=AROS_BUILD_TYPE_PERSONAL -O2 -m32 -march=i686 -Wno-unused-function -Wno-conversion-null -Wno-deprecated-declarations -Wno-cast-align -std=c++11 -O3 -DNDEBUG -fno-exceptions -fno-strict-aliasing -fno-rtti -O2 -L -static-libstdc++ CMakeFiles/owb.dir/main.cpp.obj -o ../../bin/owb ../../lib/libjavascriptcoremui.a ../../lib/libWebCoreMUI.a ../../lib/libwebkitmui.a -lz -lcairo -lpixman -lfontconfig -lfreetype2 -lxml2 -lz -lcurl -ljpeg -lpng -lssl -lcrypto -licui18n -licuuc -licudata -lsqlite3 -lxslt -ldebug -lopenurl -lrexxsyslib -lavformat -lbz2 -lavcodec -liconv -lz -lavutil -lswresample -lswscale -lm ../../lib/libWebCoreMUI.a ../../lib/libjavascriptcoremui.a /home/nicola/AROS-20181022-source/bin/linux-i386/AROS/Development/lib/libfontconfig.a /home/nicola/AROS-20181022-source/bin/linux-i386/AROS/Development/lib/libfreetype2.a /home/nicola/AROS-20181022-source/bin/linux-i386/AROS/Development/lib/libz.a ../../lib/libWTFMUI.a -lcairo -lpixman -lfontconfig -lfreetype2 -lxml2 -lcurl -ljpeg -lpng -lssl -lcrypto -licui18n -licuuc -licudata -lsqlite3 -lxslt -ldebug -lopenurl -lrexxsyslib -lavformat -lbz2 -lavcodec -liconv -lavutil -lswresample -lswscale -lpthread -lz -lstdc
There are undefined symbols in '../../bin/owb':
U CloseSocket /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavformat/udp.c:992
U connect /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavformat/udp.c:475
U recv /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavformat/udp.c:960
U send /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavformat/udp.c:981
U sendto /home/nicola/AROS-20181022-source/bin/linux-i386/gen/contrib/MultiMedia/libs/ffmpeg/ffmpeg/src/libavformat/udp.c:977
-
yes, the new ffmpeg builds by default, but the old diff is left in the source, and old ffmpeg can be compiled setting its version number in appropriate mmakefile.src. currently odyssey compiles and works (correctly?) only with the old one. to utilize new one either source needs adopted or newer acinerella needs to be pulled in and interfaced correctly, as some things changed. i have not been able to achieve this all alone.
-
@ch1ll
AFAIR bsdsocket.library is a special case. It does not physically exist (it is created at runtime) and it does not have a link library that can resolve the missing functions. Instead you need to make sure that its protos are included in every .c file that uses socket functions. These protos will redefine the function calls (like recv) into jumps relative to base so there won't be any unresolved symbols. What the error log suggest is that ffmpeg is incorrectly compiled (no protos included where needed).
-
Deadwood :D
Nice to see you again.
-
Hi all,
I was able to compile Odyssey with openssl 1.1 :) . I compiled it with ffmpeg 3.0.1, the one already available in ABI_V0 using the --disable-network flag.
I'll make some tests and share everything.
Nicola
-
Nicola: That is amazing news :) thanks a lot for your work on this. Looking forward to try it out.
-
@Deadwood
Thank you for following this and giving hints. :-)It's always a pleasure to hear you.
@Ch1ll
Keep up the good work! We absolutely need a OWB refresh, since its useability in the current web is getting worse over time. I will obviously include your updates in Icaros Desktop as soon as you'll make them available.
Kind regards
-
thank you nicola
salvatore
-
@nikos
deadwood is the guy who ported odyssey to aros and whose sources i have forked in order to get them compile foe abi v1, big endian and m68k amiga prticularly. if there is anyone to expect an informed advice from, its him.
-
Hi everybody,
Thanks for the support. I was not able to test anything, but I uploaded odyssey in case someone want to test it. I also uploaded the certificates to be uncompressed in Development/etc/SSL
https://drive.google.com/folderview?id=1INiXOKKY7KypDaLabvPo9KTN06W8--nS
@deadwood: you did an amazing job
@wava: Thanks for your help
Please let me know any feedback
-
Thanks for the link.
Tried it and program starts. Problem is with most pages I get this error.
Error 1: unsupported protocol
-
@ch1ll
Thanks, but I was just the last guy in the chain of people who brought WebKit to Amiga-like OSes. AROS version is building on top of their contributions.
BTW: If you want to merge your changes into the main line, I can give you write access to GitHub repository (https://github.com/deadwood-pl/OdysseyWebBrowser)
-
@Nicola
Thanks for the link.
Tried it and program starts. Problem is with most pages I get this error.
Error 1: unsupported protocol
To be more precise, all sites that default to an https page don't work, and give the unsupported protocol error.
pages that can actually be reached with the http (not 's') protocolo, instead, correctly work.
I have placed the ssl contents in development:etc/ssl, and replaced all original OWB directory with the one in the bigger archive.
Probably we're just missing some little step you've performed.
-
@Nicola
I done this with sashimi and snoopy to get a error log: 1. open owb, 2. open libero.it, 3. have the error appear, 4. close owb.
-
On the web, you will see two different types of URLs or links – one which starts with HTTP and the other which starts with HTTPS. The HTTPS links use SSL certificates to make the connection secure or private.
I tought SSL where part of Odyssey exe. Even if it would be prefered if it was not. I don't think Odyssey is looking for SSL in the Development path?
-
off the top of my head ssl should be linked statically right now.
edit: yes, there is pretty huge libssl.a in developer/lib
im certain there is material there to prepare interface for a shared one, but im not experienced at it. needs a .config file to define function interface and linrary init/exit stuff afair.
-
BTW: If you want to merge your changes into the main line, I can give you write access to GitHub repository (https://github.com/deadwood-pl/OdysseyWebBrowser)
might be worth to merge up some of my changes even if not necessarily 1:1 if we want to keep abi v0 compatibility. i have omitted a few workarounds you have introduced for instance due to lacking stdc headers, kalamatee implemented while i was working on odyssey compilation within aros build system. im not sure if it works with abi v0. dependant on this portions of the code could be removed or ifdeffed detecting abi v0.
there are some further things, i have updated to the latest ffmpeg version, however acinerella doesnt compile with it as it is. the solution is either to degrade to the latest ffmpeg (the diff is still in aros sources) or update acinerella adopting its interface. the current interface doesnt provide one private function any more odyssey side is dependent on. i have noticed that piru has joined in developing acinerella, but morphos team is not keen to share their patches as of current.
-
In Developement:/etc/SSL are the openssl default certificates to be trusted, and off course anything is statically linked. I guess that also Odyssey trust them. I'll give a look to this as soon as find some time.
@wava: I definitely would like to merge our source.
@deadwood: Thanks for this offer. I'll clean something and prepare a push request.
-
im not sure if krzysztof/deadwood prefers to merge stuff after pull request or rather wants others to work directly on his repo. my assumption was rather the latter, just he wants someone knowledgable and responsible enough.
what concerns my changes besides what i mentioned, there is aros mmakefile and several detailed commits to make odyssey compile with newer compilers and for other archs. i dont know how muchof this you have already integrated, ch1ll, but we will manage somehow. i guess.
-
I just run the first tests, and I found that:
1. owb every time crash when started from shell
2. the certificates of Root CA are stored in owb directory, not in development:/etc/ssl
3. I also have the same error when connecting to https sites
4. openssl command line works properly and can connect with TLS (tested with: openssl s_client -connect www.google.com:443)
It looks that there is something wrong with libcurl that was compiled without ssl support. My understanding is that the configure script test for some functions that are not there anymore.
I'll work on it and let you know...
-
If you run from shell you must give Odyssey enough stack size. The stack size used are shown in the Odyssey icon.
before running odyssey form shell type. "stack size 1192000"
-
@nikos: Yes, you are right, thanks :) .
I'm currently patching curl to make it use of OpenSSL 1.1. After that, I can link it again with owb and we can test it again.
-
Looking forward to that :)
Thanks for your work. You are realy good at this :)
I can not even compile "hallo world" ;)
-
Hi everybody,
I finally updated curl and linked owb again. You can download it on the same link: https://drive.google.com/drive/folders/1INiXOKKY7KypDaLabvPo9KTN06W8--nS
I tested it with google, repubblica.it, libero.it and something else.It looks to work. It should be tested with some misconfigured server and insecure servers.
Happy testing :)
@daedwood,wawa: I should share my changes on Odyssey and AROS ABI_V0. I modified openssl, curl and something on ffmpeg. Basing on your preference I can write to github repository, make push request or commit on svn.
I also would like to share a Vagrant file to prepare the development environment and write some instructions.
-
@ch1ll
My preference is to give you write access to github repo and have you commit changes there directly :) Can you PM me your github user?
-
Fantastic!! You made it :)
Streaming video does not work. It crash. I get software failture when Odyssey is trying to play any video.
-
@ch1ll
Great! You are doing a fantastic job! Though I have no time to test at the moment, I am glad about the progress. Keep up the good work! :)
-
Gmail works now :) Just rem. to spoof as other browser like Opera 12 and select basic view.
-
thank you very much
salvatore
-
@ch1ll
Streaming video does not work.
what ffmpeg are ybou building/linking against?
3.0.1 or 4.0.2? the latter is currently default but odyssey doesnt interface correctly to it. work on acinerella is necessary, as i already said.
if so, you can downgrade to 3.0.1 changing the version number in mmakefile.
if you keep your commits verbose ill try to keep my repo up to date and consistent with deadwoods as far as its possible for abi v1 so that we have a consistent source. i dont know exactly yet how this is done though.
-
@wawa: I used ffmpeg 3.0.1... :(
@nikos: Thanks for testing!
-
Streaming never worked good anyway. So it is not a big deal.
It is for sure maybe the most important task to solve for AROS right now, but guess not easy.
A problem that been always is that some big pages with lot's of pictures don't load them scrolling down. Try this one www.vg.no and scroll down and you see what I mean.
Any idea why that happen?
-
I don't know if I should spend time to make ffmpeg 3.0.1 to work or directly on 4.0.2 helping wawa...
I compiled ffmpeg with these additional flags:
--disable-inline-asm This was required to make it compile. I don't have anymore the error I had without it, but I found that solution on ffmpeg mailing list
--disable-network I added this to fix the issue related to owb linking issue posted before on this forum. My understanding was that ffmpeg does not manage directly connections...
If somebody has some idea let me know...
-
odyssey is currently the only program directly depending on ffmpeg. so my guess is, it would be the best to try and get 4.0.2 working with it. actually (afair) its only one function that acinerella is lacking, that doesnt get exposed by ffmpeg any more, and they claim it should have never been. on the other hand acinerella got updated in the mentime by morphos developer (piru), the sources are public. but it requires in turn further changes in at least one file where odyssey media player interfaces to acinerella, and according to andree, whom i asked, they aint inclined to share these changes, at least for now, because apparently it is not working as expected.
-
As you might know we got the sources to Odyssey web browser from a MorphOS developer. One difference with MorphOS compared to AROS is that they got overylay support in the OS. I know that Odyssey for MorphOS take advantage of that, and can for example go fullscreen, something that never worked in AROS version. The reason I say this is that there could be some code from MorphOS version that mess things up cause of this.
I know MorphOS never had the synch issues we have. No matter what ffmpeg version.
Playing video with external MPlayer there are no synch issues. What I been thinking is that there could be something inside Odyssey playing video that eat all the CPU power. I know even with Linux playing video in web-browser without GPU support can be heavy task. MorphOS is PPC only and maybe there are some things to look at. I'm thinking x86 could take advantage of some Intel CPU media extentions that PPC does not have. As you know I'm not a developer so don't laugh if this makes no scence :D
-
@ch1:
How about updating the rendering engine to a newer version ?
-
thats a whole lot of work. i have looked at the changes since odyssey snapshot and its definitely beyond what i can do. there has been structural changes for what i remember. and also i expect the current engine to be heavier. as odyssey is heavier than ten years earlier aros owb webkit.
maybe it can be done, but it needs a skilled professional, not some noob. but then, first things first.
-
@wawa:
I'm preparing the patches to ABI_V0.
I see that the code compiled by the nightly build is on the svn branch: ABI_V0-20161228 and this is not in synch in github with the repository from ezrec. Is that correct?
Since I do not have write access to svn, if you agree I can check that new openssl do not break other contribs, and then prepare a patch so you can commit it.
Regarding other developments: currently, I see as first priority to be able to have a clean compilation of Odyssey. I should be able to release it quite soon. After that, we should stabilize Odyssey and integrate new ffmpeg. If we are able to attract more developers we may think about other changes more complex like new engines.
-
@wawa:
I'm preparing the patches to ABI_V0.
I see that the code compiled by the nightly build is on the svn branch: ABI_V0-20161228 and this is not in synch in github with the repository from ezrec. Is that correct?
That is correct. ABI_V0 was abandonded ages ago in favour of the 2016 branch (which ought to be dead as well by now, but current AROS distro's rely on it)
Since I do not have write access to svn, if you agree I can check that new openssl do not break other contribs, and then prepare a patch so you can commit it.
In case you do not have write access to svn then the method you mentioned is the suggested path. You can then either post the patch somewhere, preferably on the developers mailinglist (so that it can be reviewed). Of course if wawa or someone else is helping you then a private mail to your contact is probably sane as well.
But.... did deadwood not offfer write access to his personal repository on github ? Would that not be an easier path ? Not trying to stirr things up there, just an observation.
fwiw and afaik: ezrec put up his mirror before the 2016 break and so his scripts probably did/do not include this branch (at least i was unable to find it). You can at least get a copy of the branch using svn guest account. More info also on trac.aros.org (https://trac.aros.org/trac/browser/AROS/branches/ABI_V0-20161228 (https://trac.aros.org/trac/browser/AROS/branches/ABI_V0-20161228))
-
Thanks for the suggestion. I have write access to daedwood's repository with odyssey code, but I also modified mmakefile.src for openssl, ffmpeg and for curl. On curl I also backported some function from openssl.c from the last version.
Of course I was able to access svn using guest. I will share a patch here and in the mailing list.
-
thats cool.
i can test your patches for abi v1 platforms.
if you give me your mail via pm ill invite you to
https://amigaports.slack.com
where two of main aros develpers are hanging around.
maybe also neil and deadwood are interested. or whoever else?
that not to circumnavigate dev ml, but its good for coordination, quick advice and reference.
-
btw. of course ideally at some point would be good if you gain write access to svn yourself, after neil and perhaps others approve this.
-
it seems to me that with the current engine you can use the program, wawa long ago had ported to abiv1 and nikos had tested it successfully
hi :)
-
@ch1ll:
ah yes. That would make things a bit more difficult then.
Thank you for your efforts
-
@ch1:
Consider compiling Odyssey also for abi v1_64 bit, which is the main branch; abi v0 is becoming obsolete and we should gradually move to abi v1.
-
I think it's still some work on abiv1, so this porting on abiv0 is important, even icaros desktop is based on this last branch
salvatore
-
Hi all
I finally tested again the compilation process and committed the patch to ABI_V0-20161228-patch and to Odyssey on deadwood repository on: https://github.com/deadwood-pl/OdysseyWebBrowser/tree/odyssey-r187682.
I also added my instructions on how I was able to compile.
@AROS committers: please review the AROS patches on: https://github.com/deadwood-pl/OdysseyWebBrowser/tree/odyssey-r187682/ABI_V0-20161228-patch
Basically to make Odyssey with new ssl working I had to:
- Backport OpenSSL 1.0.1
- Modify curl to make it work with new openssl
- Modify some compilation flags on ffmpeg, add AROS headers on networking files and disable udp (temporary)
Next step will be:
- Integrate with new ffmpeg
- fix Bookmark and Cookie broken dialog windows (I guess the known issue with tree list).
Please let me know your comments...
-
i have made one observation on aros dev ml. i have currently mostly cleaned my odyssey source against my github repo and am rebuilding odyssey in order to have a last state binary before ill try to apply your changes to dependencies. this way i might be able to test compare them under aros i386 abiv1 on vmware. if you have any particular test cases where there were problems that have been fixed with these changes i would be interested to know about.
so far i understand you are compiling against icu4.2.1. ill stay with 60.1 for now, as it is my default, but i can also downgrade.
and since you are already using gcc-4.8.3 maybe it would be worth to check if it also works with 5.3.0, i case the latter allows the compilation of newer icu?
-
Hi wawa
Thanks for your comment on mailing list. I modified the code on curl in order to be able to compile with ABI_V0 and openssl 1.0.1 and ffmpeg to compile with networking. On openssl I used the same patch from ABI_V1 and then I modified the mmakefiles.
It should compile with other version of gcc, but some small change are required. For example you need to remove the -newclib flag. I never tested other versions of icu.
Please let me know how it goes for you...
Last but not least: if you make a push request with your changes on Big Endian I'll test the compilation and integrate in daedwood repository
-
@ch1ll
so i have now built the odyssey again and left you message on slack aros channel about how i intend to proceed.
-
for the moment that odissey has problems, you can solve some problems by going to this address http://pici.picidae.net, in the page that opens indicate the site you want to reach, like facebook, ebay etc ..
hi
Salvatore
-
Yes Salvatore it works well though "pici.picidae.net" released a long time ago it served for other things; on OS3 even the old AWeb and Ibrowser can see on modern Web pages thanks to "pici.picidae.net", i talked about it HERE (https://amiganews.it/forum/viewtopic.php?f=4&t=19523&start=30)
-
for the moment that odissey has problems, you can solve some problems by going to this address http://pici.picidae.net, in the page that opens indicate the site you want to reach, like facebook, ebay etc ..
hi
Salvatore
Did not know about that :) Very handy.
Thx Salvatore.
Even using that site there are some sites like www.vg.no that does not display images scrolling down the site. Anyone know why?
-
The site www.vg.no both works for me on OWB and on NetSurf, AWeb and IBrowser using "pici.picidae.net", but it works well even with OWB only.
@Admin
There are problems to send a message, many times waiting times are very long (maybe this happens when some other user is connected to the same discussion ?)
-
hello I'm glad you're happy we hope the service continues to exist, I discovered this thing by chance is a retroaccademy article :)
hi
salvatore
-
The site www.vg.no both works for me on OWB and on NetSurf, AWeb and IBrowser using "pici.picidae.net", but it works well even with OWB only.
If I turn of java-script it is like you say working well. Problem then is that many parts of the site is not displayed. For sure something happening regarding java-script. I think java-script might be the problem also with youtube.
-
Hi all
I finally tested again the compilation process and committed the patch to ABI_V0-20161228-patch and to Odyssey on deadwood repository on: https://github.com/deadwood-pl/OdysseyWebBrowser/tree/odyssey-r187682 (https://github.com/deadwood-pl/OdysseyWebBrowser/tree/odyssey-r187682).
I also added my instructions on how I was able to compile.
@AROS committers: please review the AROS patches on: https://github.com/deadwood-pl/OdysseyWebBrowser/tree/odyssey-r187682/ABI_V0-20161228-patch (https://github.com/deadwood-pl/OdysseyWebBrowser/tree/odyssey-r187682/ABI_V0-20161228-patch)
Basically to make Odyssey with new ssl working I had to:
- Backport OpenSSL 1.0.1
- Modify curl to make it work with new openssl
- Modify some compilation flags on ffmpeg, add AROS headers on networking files and disable udp (temporary)
Next step will be:
- Integrate with new ffmpeg
- fix Bookmark and Cookie broken dialog windows (I guess the known issue with tree list).
Please let me know your comments...
Hi. I'll start with a big thank you, since your build of OWB seems also more stable (at least, this is the feeling I have today with current websites). The only issue I have is that its executable is around 100 MB bigger, and does not fit at all in Icaros Desktop Light. Is there any plan to shrink/strip it down, maybe using some development magic I am totally unaware of?
If not, just tell me and I'll try to reconsider some of Icaros Light contents in order to make the necessary room.
-
So good news I would say, thank you very much ;)
-
Hi. I'll start with a big thank you, since your build of OWB seems also more stable (at least, this is the feeling I have today with current websites). The only issue I have is that its executable is around 100 MB bigger, and does not fit at all in Icaros Desktop Light. Is there any plan to shrink/strip it down, maybe using some development magic I am totally unaware of?
Thanks for testing. I can try to strip debugging symbols and disable video until I cannot make it work. I'll keep you posted.
-
I can not test the program at the moment, it seems strange to me that I can not boot on a car that is years old and they are two
salvatore
-
I downloaded icaros desktop light 1.4 everything works perfectly, when I have time I try the last of the 1.x series, I will update what I need, like the fat handler etc..etc ..
ps. Paolone I make a donation as soon as possible, because without it I could not do anything
All the best
Salvatore
-
Hi. I'll start with a big thank you, since your build of OWB seems also more stable (at least, this is the feeling I have today with current websites). The only issue I have is that its executable is around 100 MB bigger, and does not fit at all in Icaros Desktop Light. Is there any plan to shrink/strip it down, maybe using some development magic I am totally unaware of?
Thanks for testing. I can try to strip debugging symbols and disable video until I cannot make it work. I'll keep you posted.
@ ch1ll :
can you make video a separate module that can be enabled/disabled by the user ?
You could also make a shared (re-entrant) library (to be put in LIBS:) of some of the static linked libraries that owb uses; for example a 'curl ' shared library may benefit also other applications who need to fetch something from the internet.
-
I thought I could use this version of owb with icaros 1.4 is the only part of my computer imho, however with a pentium 4 I should go on the safe side, I will spend another 20 euros
hi
Salvatore
-
I downloaded icaros desktop light 1.4 everything works perfectly, when I have time I try the last of the 1.x series, I will update what I need, like the fat handler etc..etc ..
Hi. The 1.x series has been deprecated for years. Too much old AROS code, and some applications *need* at least Icaros 2.0 to work.
-
Yes Paolone i verified :)
-
Hi. I'll start with a big thank you, since your build of OWB seems also more stable (at least, this is the feeling I have today with current websites). The only issue I have is that its executable is around 100 MB bigger, and does not fit at all in Icaros Desktop Light. Is there any plan to shrink/strip it down, maybe using some development magic I am totally unaware of?
You can find a new version without debugging symbols on this link: https://drive.google.com/open?id=1D_e0EXa7IYyix5GBhGYsmTqQpvn67iA-
I did not ffmpeg 3.0.1, since I found why it was crashing every time. There is a bug in AROS ABI_V0 on the posix function posix_memalign. I modified the code to not use it.
Please let me know how is it...
-
Thanks. There is something real good that happend with your new version. It is very fast!! :)
Video does not crash anymore but got playback error trying to play.
This is what I got from youtube.
An error occurred. Please try again later. (Playback ID: ZF-olCE815RfzoSj)
Learn More
-
I also have that error with youtube. This is compiled with the version 3.0.1 of ffmpeg with the old integration code.
Do you have the same error also with the previous version compiled by deadwood?
-
The old version from Deadwood youtube works, but as you know there are some issues with the playback.
-
I also have that error with youtube. This is compiled with the version 3.0.1 of ffmpeg with the old integration code.
Do you have the same error also with the previous version compiled by deadwood?
Neil will try to fix posix_memalign. Lets hope it is related to the video problem. He will not test it in Odyssey, but guess you can do that ch1||.
I keep you posted.
-
I would be cool to have the fix for posix_memalign, but I guess the issue we are facing with youtube is something else...somebody knows what version of ffmpeg was used by deadwoood?
I would not like to spend much time to debug ffmpeg3.0.1, but I'm working to make it work with ffmpeg 4.0.1...
-
I can not get into a site (imusician/login) I do not know if it's a ssl or engine protocol problem
hi
Salvatore
-
if I identify myself as opera 12 everything is fine, thanks for everything nicola
hi
Salvatore
-
Neil will try to fix posix_memalign.
Very appreciated! :D I am looking forward to the fix. Thank you for your efforts, Neil!
-
To get the streaming working at least with youtube would make our system so more usefull.
The latest compile of Odyssey with new SSL is very nice. Fast and quite stable. Not that many pages not working. I would say at least 80% works as in any other browser.
I'm thinking if it is possible to make a post at amigaworld and point to the source code where the video stuff are. There are many talented developers looking in forums that could at least come up with some ideas of what is the reason for this. Also some good explenations about what happen trying to play video are important.
If Ch 1|| can get new ffmpeg working with Odyssey it might solve the problem. I'm not sure what ffmpeg version MorphOS use.
First of all this should be a straigh port from MorphOS version where Video does work.
What I know about MorphOS is that they have overlay and we don't. That should not have
anything to do with it I guess.
-
Neil will try to fix posix_memalign.
Very appreciated! :D I am looking forward to the fix. Thank you for your efforts, Neil!
Is this already fixed? I am just curious. :)
-
First of all this should be a straigh port from MorphOS version where Video does work.
What I know about MorphOS is that they have overlay and we don't. That should not have
anything to do with it I guess.
Not quite. The embedded video player did not come from MorphOS iirc.
-
Ah, thx terminills. Did not know that.
-
I did not ffmpeg 3.0.1, since I found why it was crashing every time. There is a bug in AROS ABI_V0 on the posix function posix_memalign. I modified the code to not use it.
Please let me know how is it...
What's the bug in posix_memalign()? I extended the test program for it, but couldn't get it to fail except when there wasn't enough memory available. I tested it on ABIv0 on pc-i386 and darwin-i386 (and also on ABIv1).
The new test program is only committed to SVN for ABIv1 for now, but I can backport it if there is interest.
-
Hi all,
After some time I wanted to work on Odyssey, and I compiled for both ABI_V0 or ABI_V1.0
Compilation is fine, it start, but when I try to open any website I got an error "Unknown CPU Error" ... "Wait for disk the activity to finish".
I never seen this error before. Any ideas?
-
null pointer when using OM_SET metod with listtree, maybe its wrong that it is called for third time in a row, where it obviously gets a wrong address (deado15c) or runs off allocated memory or whatever the correct answer to that is. because im a noob i usually dont get bugs like that sorted as well. better you ask someone competent on our slack.
-
you need to look up how dispatcher stuff with mui/boopsi works.
for amiga you might look somewhere around here: http://amigadev.elowar.com/read/ADCD_2.1/Libraries_Manual_guide/node0213.html
aros has also a document about that:
https://en.wikibooks.org/wiki/Aros/Developer/Zune/Classes
further down that page.
-
Hi Wawa
Thanks for the tips! I'll give a look to it and I will ask on slax if I still need help.
I was looking to some link like this since long time :D
-
@ch1ll
thank you for effort :)
-
Good news. Would be great if streaming could work as it should.
-
hi all
practically there are other services accessible with the current version of owb for example, I can't access my gdrive or open a DropBox account, but I think you know better than me, I noticed the situation NG also Morphos and AmigaOS4 have this gap, but the latter will have the full support of ibrowse 3.0, when it comes out and I think soon, it is a pity it is so hellish to work with the code regarding our flagship application
All The Best
-
I discovered that if you use navigation as windows 10 compatibility, the artifacts on the screen disappear, I think it's a good find, it was really annoying
hi all
-
While reading this thread I noticed someone mentioning that it's possible to compile for ABI_v0 or ABI_v1.0 I assume that means GCC.
I'd like some assistance or direction to resources to set up the GCC Toolchain to compile for ABI_v0 for Icaros on Linux 32bit. Any help is appreciated.
My ShowPicture graphics program is close to first release. I have some valuable code that would help with datatypes. I would like to compile test and troubleshoot all of it before releasing it.
-
Hi miker1264
Here how I compile AROS for ABI_V0 with gcc-4.8.3. It can be compiled also with gcc-6.1.0 but you need to remove -newclib flag from few files before the configure command.
In order to compile Odyssey few patches are required. If someone is interested I can write more details.
GCC_VERSION="--with-gcc-version=4.8.3"
SOURCE="ABI_V0-20161228"
TARGET_DIR=linux-i386
svn co https://svn.aros.org/svn/aros/branches/ABI_V0-20161228
mkdir $TARGET_DIR
cd $TARGET_DIR
../$SOURCE/AROS/configure --enable-ccache --target=linux-i386 --with-portssources=~/PortSources --with-paranoia=no $GCC_VERSION --with-binutils-version=2.25 2>&1 |tee configure.log
make -j3 2>&1 | tee compile-aros.log
P.S. The problem I posted before is solved. More details on AROS-DEV Mailining list. I'll post it also on slax.
-
Thank you. That helps quite a bit.
I'll try to instal GCC 4.8.3 sometime this weekend.
I haven't used GCC before so it will be an experiment at first. I read somewhere that AROS includes are on the install disk. Are there more updated includes or is that what you use for compiling Odyssey? Would this same compile sequence work for Magellan and MultiView as well as datatypes as long as they are in the sources?
-
Here how I compile AROS for ABI_V0 with gcc-4.8.3. It can be compiled also with gcc-6.1.0 but you need to remove -newclib flag from few files before the configure command.
I updated the patch for ABIv0 GCC 6.1.0 to support the -newclib flag:
https://ae.amigalife.org/index.php?topic=60.msg1381#msg1381
With this I was able to compile the AROS with it without poking the configure files.
-
Hi miker1264
Here how I compile AROS for ABI_V0 with gcc-4.8.3. It can be compiled also with gcc-6.1.0 but you need to remove -newclib flag from few files before the configure command.
In order to compile Odyssey few patches are required. If someone is interested I can write more details.
GCC_VERSION="--with-gcc-version=4.8.3"
SOURCE="ABI_V0-20161228"
TARGET_DIR=linux-i386
svn co https://svn.aros.org/svn/aros/branches/ABI_V0-20161228
mkdir $TARGET_DIR
cd $TARGET_DIR
../$SOURCE/AROS/configure --enable-ccache --target=linux-i386 --with-portssources=~/PortSources --with-paranoia=no $GCC_VERSION --with-binutils-version=2.25 2>&1 |tee configure.log
make -j3 2>&1 | tee compile-aros.log
P.S. The problem I posted before is solved. More details on AROS-DEV Mailining list. I'll post it also on slax.
So you will soon release a new build?
Is the youtube, streaming problem fixed too?
Thank you for your work.
-
thank you very much guys
-
@ ch1ll :
Can you tell us what are you doing with Odyssey right now ?
-
I updated the patch for ABIv0 GCC 6.1.0 to support the -newclib flag:
https://ae.amigalife.org/index.php?topic=60.msg1381#msg1381
With this I was able to compile the AROS with it without poking the configure files.
Thank you very much! It works :)
-
@ ch1ll :
Can you tell us what are you doing with Odyssey right now ?
As of now I rebuilt my dev environment and cleaned up my fixes.
Then I need to find an efficient way to debug... as of now the link of owb with debug symbols fails.
@nikos: I just restarted the work on Odyssey...
-
great ch1ll :D
-
Then I need to find an efficient way to debug... as of now the link of owb with debug symbols fails.
I was never sucesfull with this. The debug binary is about 650MB in size and it failed to load under linux-hosted. The only way to debug was via printing messages via bug().
-
hopefully well, but it seems to me that the most work is on ffmpeg
-
surely having the browser and view the pages without artifacts is something else, I was really getting tired ;)
-
ebay and facebook don't show image item/post, maybe there is the problem with ssl library
Best