Blogs

Haiku monthly activity report - 03/2015

Blog post by PulkoMandy on Sun, 2015-03-29 12:55

Hello there, here comes the activity report for the month of march 2015.

This month there were 104 commits (hrev48848-hrev48952), 5 more than in the previous month.

General news

First of all, some news not directly related to the commits which happened this month.

As there was no release of Haiku for more than 2 years since Alpha 4, people at TuneTracker Systems decided to release their own distribution of Haiku. It comes as an USB stick you can order at http://discoverhaiku.com/ . TuneTracker Systems is one of the profesional users of Haiku, and they contribute to the project by providing a lot of testing and feedback, donate money to Haiku, inc. whenever possible (all their computer purchases are done using goodshop, for example), and pay developers directly when they really need a problem solved (this year they paid mmlr to work on the CDDA issues).

izcorp (http://www.izcorp.com/) is another company is planning to use Haiku in a commercial product. Their line of studio recording systems is currently running BeOS and Zeta, but they are working on an update to Haiku. Ithamar is working with them to get their hardware fully supported, and the changes will be upstreamed to Haiku in the coming weeks. This includes several fixes to the USB stack, the intel_extreme driver, and there could be more to come.

We have also reached the mark of 50 tickets left before beta1 happens.

User interface

Waddlesplash continued his efforts on cleaning up the Tracker code and fixing several minor problems with it. It is nice to see someone taking care of Tracker and cleaning up the code. While there aren't much functional changes yet, having
a clean code base is important for future improvements. John Scipione also helped with this, and cleanups in various Tracker add-ons, Deskbar, and the storage kit.

John also fixes DiskUsage to give more useful results, as it was a bit confused by the package filesystem. This also included some minor ui and code style fixes. He also improved the look of radio buttons and checkboxes with bigger font sizes.

Janus contributed several user interface fixes. Showimage, mail preferences, appearance, printers, media, got small interface changes. The goal is to make the apps all use the same conventions, for example the same positionning and ordering of default/revert buttons, and some behavior fixes (disabling the buttons as appropriate).

Axel merged his work on the network preferences. Most of the issues with the new preflet are now fixed (there are some left), and the interface between the app and the network kit was cleaned, which could make things simpler for alternative network managers as well. This work also includes a more complete "services" tab allowing to configure SSH and FTP. This is one more feature from the BeOS we were still missing.

Jessicah and waddlesplash made some small improvements to pkgman console output. The progress bar will work properly when the terminal is less than 80 chars wide, and it will show "update" instead of an "install/remove" pair when updating a package.

New functionality

3deyes contributed a translator for DejaVu files, as well as a viewer for the DjVu format.

Packages

Korli continued his work on outsourcing packages from Haiku sources. We are now using external packages for zip, netcat. Existing packages for wget, dtc, lzo, libssh2, mpg123, harfbuzz, nasm, libvorbis, openssl, libxml2, llvm, nano, icu, libpng were also updated.

Humdinger also contributed some packages (additions and updates) for Haiku native apps and ports: yab, unrar, befar, beae, weather, beezer, desknotes, documentviewer, MeV, seeker, slayer.

Joe Prostko added packages for Fossil, which are the result of a long term effort to get the tool running on Haiku and upstreaming the changes to Fossil.

Waddlesplash added packages for LMMS, a Qt-based media player and music library manager.

Drivers and kernel

Jérôme Leveque contributed patches to fix the ice1712 sound card driver which would play garbage on system with more than 256MB of RAM. The hardware can only address the first 256MB of memory, and the driver would not make sure the buffer were allocated in that area. It does now, so the correct data is sent to the souncard.

Korli also worked on some improvements to the AHCI (SATA hard disks) support, which still has problems on some machines. He also fixed an issue in our C library which would trigger a crash when using threads with libroot_debug.

A series of patches from Barrett were pushed, which improves several aspects of the Media Kit. Besides the usual round of cleanups, this improves the shutdown/restart process of the media servers, and includes a rework of the BTimeSource class. There are more patches being reviewed on Trac.

Axel added support for B_MIME_TYPE in listattr and addattr, so you can now use these to manipulate the mime type of files.

mmlr fixed a problem in the rootfs where renaming an entry to a longer name would overflow a buffer and corrupt memory. This was probably the cause for at least some of our CDDA issues. This was a longstanding problem for TuneTracker Systems, as they use CDDA a lot.

Kallisti5 started work on Raspberry Pi 2 support. The first version of the Pi was hugely popular, and one of the devices that people wanted to run the ARM port of Haiku on. However, it used an ARMv6 CPU, while every other modern system uses version 7 (or 8). The support for ARMv6 needs some extra work in the kernel, which wasn't worth the effort, even for a target as popular as the Pi. But the new Raspberry Pi 2 comes with an ARMv7 CPU. This means we will be able to use the same ARMv7 system on it as on all other ARM devices. This makes it a much more interesting target. The work done involves some cleanup of the existing Raspberry Pi support (since some of the hardware is similar, that part of the work is not lost), as well as general work on FDT support (the FDT is a way to get the same kernel to run on all ARMv7 devices, like we do for x86, instead of having a specific kernel for each board with hardcoded hardware support). Pawel Dziepak also helped improve some of the drivers for the Raspberry Pi 2, using C++14 features to make the code simpler (we can do this on the ARM code since there is no gcc2 legacy there). mmu_man also helped untangle the u-boot entry points (we still have some code left from his early u-boot experiments there, not all of which is useful with more recent versions of u-boot).

Haiku monthly activity report - 02/2015

Blog post by PulkoMandy on Mon, 2015-03-02 09:38

Hello there!

My contract has ended, but for now I have some free time to write a report every month about the ongoing development efforts from the Haiku team. I think this is a nice way to better see the work done, more so than looking at the roadmap progress bars which tend to not move much.

This month there were 91 commits (hrev48757-hrev48848). Let's see what's inside those.

End of contract - closing words.

Blog post by PulkoMandy on Wed, 2015-02-18 08:21

Hi,

As you probably have noticed, there were no weekly report in the previous weeks. The reason for this is that my contract is currently stopped. There is currently not enough money in Haiku's treasure chest to safely continue it. So, it's time to me to get back to "real life" and a full-time job in a software development company.

First of all I want to thank everyone who made this long contract possible by donating money to Haiku. It was a great experience for me, and a lot of fun as well. I did my best to move Haiku forward towards the R1 release. Unfortunately the beta 1 still isn't there, and we currently have 57 blocking tickets. It is a small number, but only the most complex or big issues are left.

Google Code-In 2014 Wrap Up Report

Blog post by scottmc on Fri, 2015-02-13 08:53

Google has now announced the 24 winners for Google Code-In 2014, with Josef Gajdusek and Puck Meerburg being the two winners from Haiku. This is Puck's second time winning for Haiku. This year we got to pick our top 5 out of the top 10 students who completed that most tasks for Haiku. Augustin Cavalier was selected as our backup winner, and Markus Himmel and Chirayu Desai were selected as finalist. Chirayu was a GCI 2013 winner with RTEMS, and made the jump to Haiku when RTEMS took this year off from GCI.

This was the fifth year of Google's Code-In, and the fifth for Haiku. This year we had 6 students who completed 20 or more tasks, one more than in 2013. We had 36 students who completed three or more tasks and qualified for a Google Code-In T-Shirt, and 53 students who completed two or more tasks. This was the first year of having beginner tasks, aimed at lowering the bar to get more new students introduced into open source. Haiku had 149 total students complete at least one task, many of those were for the beginner tasks. We had 164 beginner tasks completed, which was mostly just to introduce students to booting and using Haiku. Other beginner tasks were to compile Haiku or to install and use Haikuporter to build a package from a recipe file. In total students completed a staggering 435 tasks this time for Haiku.

Contract weekly report #61

Blog post by PulkoMandy on Fri, 2015-01-30 09:27

Hello there!

As you may have noticed if you watch the commit list closely, my libbind work has not been merged yet. There are still some bugs to solve there, but I got sidetracked. I use BReferenceable in my DNS cache implementation to keep track of the cache entries. BReferenceable is a class used in Haiku to implement reference counted objects. In C++, the language only has very simple memory management, in the form of the new and delete operators. Objects can be allocated on the stack (they are temporary and only last as long as the function they are declared in is executing), or on the heap (for long lived objects). Objects allocated on the stack are deleted automatically when the function exits, while objects allocated on the heap must be deleted manually. This is one of the annoying parts of C++: managing the lifetime of these objects, making sure they are deleted only once, and that no one will try to use them after deletion.

Tracker-Layout has landed!

Blog post by waddlesplash on Tue, 2015-01-27 22:00
tracker_layout — the (semi)famous branch that rewrites Tracker to use the Layout Kit — has been merged.

Contract weekly report #60

Blog post by PulkoMandy on Fri, 2015-01-23 07:39

Hello world!

Not much commits from me this week, as I'm still working on the libbind update, and I'm also doing some work for other customers. I got netresolv to build after implementing the missing getifaddrs function in Haiku - this is a non-POSIX function, but it is available in Linux and all major BSDs. It enumerates all network addresses for all network interfaces on the system, similar to our BNetworkRoster and BNetworkInterface classes.

Syndicate content