Odyssey 1.23: AOS4 source code on GitHub

@kas1e

Got it build on Linux. Not tested yet.

There was a strange issue: during compilation "BCRefPtrWTF.h" couldn't be found as if some include path was missing. This file exists in BAL/Types/WTF. I just ended up commenting it out for now. It was only included in one place.

Second small issue some other Linux user might bump into: https://github.com/kas1e/Odyssey/blob/ ... O_BUILD_AMIGAOS4.txt#L162 "OwbLauncher" after -o should be written like "OWBLauncher".

@Capehill
Issue with that include should be harmless. I assume it can be the same "big/small letters" issue somewhere (or in the file which do include it, or in the name of file, etc).

Are your linux box modern enough? How fast odyssey builds on?

Hi All,

I noticed there are no issues raised on the github page. How do you want to track bugs? I've been using Odyssey for many years but have not been posting the crashlogs anyplace. Does it make sense to start to create issues for items like that?

Also, big thank you for working on this project! I know I speak for a lot of OS4 users who are existed to see this development continue!

Cheers,
BIll "tekmage" Borsari

Join me in Silicon Valley June 6th and 7th for www.amiga35.us!

@tekmage
Most of bug-reports and suggestions were there : http://bugs.os4depot.net/?function=viewissues&softwareid=15

But it can be good idea to use github tickets now because on github it all just easer and better to discuss.

@kas1e

Quote:

There is also that ".exe" part in HOWTO which doesn't apply to Linux.

Quote:

My laptop is some years old, it has an I7 + SSD. I have now timed the build twice and it takes about 38 minutes with -j4.

Quote:

Pretty fast! On cygwin on the same sort of notebook as for you, it probabaly take not 1 hour as i think at first, but 1.5-2 hours i think. That all because of this "cmake" scaning on dependencies and calling fork() for, which on cygwin mean new processes and stuff as it can't be done like real unix's fork()

@kas1e

Just adding some notes about my setup:

Using Linux Mint 19.3.

Adtools:

1) built using instructions https://github.com/sba1/adtools
2) Needed to change the SDK version as mentioned in https://github.com/sba1/adtools/issues/60 and
3) GCC version as mentioned in https://github.com/sba1/adtools/issues/81
4) Also created a symbolic link from /usr/local/amiga to native-build/cross-root/bin directory.

Odyssey:

1) building as "su". Have to check again can it be done as an ordinary user.
2) Comment out one include "BCRefPtrWTF.h" as a workaround. To be checked.
3) At linking phase I'm using ppc-amigaos-g++ and removed -lstdc++. To be checked which way is better.

@Capehill
Quote:

That one will surely harmless. I remember i add it myself for Cygwin. Probabaly something somewhere about big/small letters still.


Quote:

gcc with -lstdc++ and g++ without, produce the same size of binary, and it works the same. But probably you right if you mean to make it all looks cleaner, less "-l" better at least visually )

@kas1e

I did some work on build system to reduce manual steps. Now building and linking should work when running make.

https://github.com/capehill/Odyssey/tree/fix-cmakefile

Oops. Now I see that I'm using -athread=native instead of -athread=single. Could change that.

In addition to that, I have a build.sh script that

1) runs cmake with all those switches, and
2) runs make -j4 owb.

@Capehill

Thank you for your work!

@Capehill
Thanks, will merge ! (update your repo too, fixed that av_free() crash)

@Capehill
Added your changes: now all builds till the end. I also change "owb" on "Odyssey" binary, so now you run "make -f Makefile Odyssey", and binary will be in the build/bin directory.

Also adapted how_to_build.txt accordingly.

I've used Odyssey for years on my XE and X5000 as my daily browser without problem, by choice if you wonder why!

This new developement is fantastic,
Oddysey was fast before, but these
new betas is really kicking some shit!

@kas1e

Confirmed. Lazy people can just do "make -j4 Odyssey" (no need for -f Makefile).

@Capehill

Sorry for the "not so on topic" question (It's a new hobby of mine, it seems) :-/

Quote:
But why???
I was under the impression that -athread=native is the way to go when compiling stuff for AmigaOS4?
Why single now?

Has something changed in the meantime that i missed?
Doesn't single mean that there won't be concurring/parallel threads...why use pthreads at all then?

I'm such a noob

@Raziel
You need to create topic called "Raziel's questions" and put all there :)

pthreads with odyssey used not because of athread-single, but because of some other 3d party libraries which use pthreads. You can buld odyssey and with -athread=native, but it will make no since, only binary will be bigger. Application is single threaded, so no point to use athread=native.

@Raziel

I guess the OWB code we are using at the moment is older C++ (< C++11) so it doesn't use C++ concurrency features. If I remember correctly the newer Webkit uses newer standard and if we get there, this may need to change.

@kas1e However there are multiple tasks related to OWB like MediaPlayer.

@Capehill
Yes, but they are real amigaos processes created by CreateProcTags, and not handled via c++ threads. I.e. timer, mediaplayer, icon handling and some other taks are real amigaos processes created via CreateProcTags in few parst of Odyssey's code (maingly in WTF).

C++11 and stuff will help indeed only on later versions of webkit (like those revisions on which 1.25 is based, and which i port too already with some bugs which need to fix, but it indeed use c++11 threads and there athread=single does not work anymore and only athread=native can be used because of c++11 threads in the webkit itself).

But with the version we works currently athread=native just of no needs and size-increase. You can surely build it right now with athread=native, it just will give no benefits at all, only bigger size of file.

@Capehill
Made big commit today : added proper debugging functionality: defined bug(), so D(bug("...")); can be used. Thus we use DebugPrintF() on OS4 everywhere instead of kprintf() and get rid of clib2's libdebug.a. There are new file "Base/debug.h" which currently looks like this:



Also removed everywhere include of clib2/debug_protos.h (as it anyway now included for os3 and morphos from debug.h).

Rebuild everything from scratch: all seems to builds and works fine.

If you have any suggestions/ideas about plz feel free :)

@kas1e

Nice work, pulled and built successfully.

Need to start fixing those compiler warnings, too ;)