Very good job !
Two problems for me:
I don’t manage to test webgl water test, maybe due to my old hardware (ATI H4550 RV710)
Thanks for all your staff !!
I’ve tried to follow the instructions on the wiki to compile :
I’m OK until the ./mach boostrap and then with ./mach build I have the below : any idea what’s wrong ?
0:00.70 W Clobber not needed.
0:00.76 W Adding make options from /boot/home/Documents/C/gecko-dev/mozconfig
MOZ_OBJDIR=/boot/home/Documents/C/gecko-dev/obj-ff-dbg
OBJDIR=/boot/home/Documents/C/gecko-dev/obj-ff-dbg
FOUND_MOZCONFIG=/boot/home/Documents/C/gecko-dev/mozconfig
export FOUND_MOZCONFIG
0:00.76 /bin/make -f client.mk -j4 -s
0:01.53 Elapsed: 0.00s; From dist/private: Kept 0 existing; Added/updated 0; Removed 0 files and 0 directories.
0:01.54 Elapsed: 0.00s; From dist/public: Kept 0 existing; Added/updated 0; Removed 0 files and 0 directories.
0:01.72 Elapsed: 0.14s; From _tests: Kept 679 existing; Added/updated 0; Removed 0 files and 0 directories.
0:01.95 Elapsed: 0.51s; From dist/include: Kept 6561 existing; Added/updated 0; Removed 0 files and 0 directories.
0:02.24 Elapsed: 0.32s; From dist/bin: Kept 2908 existing; Added/updated 0; Removed 0 files and 0 directories.
0:02.49 ./ServoStyleConsts.h.stub
0:02.95 thread 'main' panicked at /boot/home/.cargo/registry/src/index.crates.io-6f17d22bba15001f/cbindgen-0.27.0/src/bindgen/config.rs:1125:34:
0:02.95 called `Result::unwrap()` on an `Err` value: "Couldn't parse config file: TOML parse error at line 355, column 1\n |\n355 | \"Keyframe\" = \"Keyframe\"\n | ^\nduplicate key `Keyframe` in table `export.rename`\n."
0:02.95 note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
0:03.00 backend.mk:571: recipe for target 'layout/style/.deps/ServoStyleConsts.h.stub' failed
0:03.00 make[3]: *** [layout/style/.deps/ServoStyleConsts.h.stub] Error 101
0:03.00 make[3]: *** Waiting for unfinished jobs....
0:03.18 /boot/home/Documents/C/gecko-dev/config/recurse.mk:32: recipe for target 'export' failed
0:03.18 make[2]: *** [export] Error 2
0:03.18 /boot/home/Documents/C/gecko-dev/config/rules.mk:361: recipe for target 'default' failed
0:03.18 make[1]: *** [default] Error 2
0:03.18 client.mk:59: recipe for target 'build' failed
0:03.18 make: *** [build] Error 2
0:03.20 W 0 compiler warnings present.
Welcome to the gecko hacking land!
I didn’t see that error. It seems cbindgen failed to generate .toml file probably rust copmpiler use.
cbindgen is a rust binding generator for C(/limited C++) library. Mozilla’s build system has build phases including configure, export, compile, misc and so on. In export phase it generates source files from data or other type files. cbindgen failed to generate correct output here, but not catched.
I don’t know what failed in detail, but want to see clean generated files and regenerate it.
Run ./mach clobber which cleans build output, then ./mach build again.
See also:
I just find an article for you.
Thank you for reporting. Registered issue for the former. The latter is known and some people pointed out.
OK I found the problem
cbindgen 0.27 was not compatible with our tree
to fix we can
Could you try the latter for now?
With cbindgen 0.26 indeed I dont’ have anymore the error. Compilation in progress.
Thanks 
I reach “wgpu_bindings” with the below error : any idea about the I/O failure ?
Maybe not enough remaining memory ? (90% of my 8 Gb are used)
13:28.13 Compiling fluent-langneg-ffi v0.1.0 (/boot/home/Documents/C/gecko-dev/intl/locale/rust/fluent-langneg-ffi)
13:31.27 Compiling wgpu_bindings v0.1.0 (/boot/home/Documents/C/gecko-dev/gfx/wgpu_bindings)
14:37.30 rustc-LLVM ERROR: IO failure on output stream: Permission denied
14:37.45 error: could not compile `tabs` (lib)
14:37.45 warning: build failed, waiting for other jobs to finish...
14:37.89 dom/abort
14:38.53 clang++: warning: argument unused during compilation: '-fstack-clash-protection' [-Wunused-command-line-argument]
15:34.91 dom/animation
15:35.65 clang++: warning: argument unused during compilation: '-fstack-clash-protection' [-Wunused-command-line-argument]
“Firefox” brand name and icon is probably not allowed to be used in HaikuPorts until Mozilla will have official Haiku support.
If log is correct, it fails with a permission error to output. When I start building, I sometimes experienced file system error with USB thumb drive.
What I concerned is which is your build directory a normal BFS partition on SSD/HDD, a VM image or some special partition or filesystem?
I successfully build on BFS partition of USB SSD of bare metal.
Let’s hope Mozilla will give Haiku its blessing for using the Firefox brand.
It doesn’t really matter.
Using the Nightly brand,which I think is allowed without special agreement from Mozilla,won’t make it a different browser.
On the other hand,using a non-Firefox brand will give us a lot more freedom to finetune the settings and disable quite a few antifeatures by default.
16 Gb seem to be required as it’s now ok.
It’s a beast to compile 
I’ve used “./mach build”, what are the options to make the executable lighter ?
My version compiled :
3060336 oct. 21 05:23 firefox
Your version distributed :
358432 oct. 14 18:29 /boot/home/Desktop/firefox/firefox
And it’s exiting more often due to error :
Exiting due to channel error.
You need to use haiku_mozconfig instead of haiku_mozconfig_dbg to get an optimized release build. Our build instruction specifies the former for development use.
You can change that by
rm mozconfig and ln -S haiku_mozconfig mozconfigexport MOZCONFIG=path/to/haiku_mozconfig to override mozconfig specificationThis is effectively changing following build settings
Such a big build config change results a full-build, so next ./mach build should take hours.
EDIT: you can make tgz package with ./mach package after build finished. What I uploaded was made with such steps.
You need revert this commit
I will push this to haiku128 branch later.
Done in
With the commit reverted indeed it’s more stable now it tooks only a few minutes to recompile. (Versus more than 5 hours for the whole)
Thank you for the details of the mach build, I have now a working env for compiling gecko-dev : I think 8 Gb of memory is not enough, so maybe this can be indicated on the wiki to avoid this pitfall
I have a question as I’m new to the debugging land :
Why the Haiku Debugger tool is not displaying the source of the firexfox program even if it’s compiled with debug info ?
file /boot/home/Documents/C/gecko-dev/obj-ff-dbg/dist/bin/firefox
/boot/home/Documents/C/gecko-dev/obj-ff-dbg/dist/bin/firefox: ELF 64-bit LSB shared object, x86-64, version 1 (SYSV), dynamically linked, BuildID[sha1]=2fd00db3ef61a3870fe8bbe568089e16aa74d066, with debug_info, not stripped
For a simple OpenGL program it’s working fine as per below :
But for firefox main program, the source is not displayed :
Short answer is, use GDB15 to debug firefox.
It is why I played with Haiku Native debugger. I was there last year:
There are two problem with Native Debugger:
So I tackled try porting LLDB but it didn’t succeed then by lacking both of my skill and my spare time
In this year’s GSoC, a student @trungnt2910 worked on improving this situation
and he did! He ported latest GDB again and made it very stable enough to debug the firefox.
His blog series on this is great to read.
It can be used to debug C++/Rust mixed project. I’m using Qt Creator for GDB graphical frontend.
EDIT: I wrote GDB as GCC, I was sleeping in bed then.
This is my working debug configuration of Qt Creator
-no-remote -profile /boot/home/src/gecko-dev/obj-ff-dbg/tmp/profile-default./mach run./mach run executed inIf you didn’t export WAYLAND_DISPLAY=:0 when you launch Qt Creator, you need Manage the Kit and specify that in Environment: field.
