Every other week, someone asks how to use the new unit testing framework in GLib (released with GLib-2.16.0) and Gtk+ (to be released with the next stable). First, here is a write-up from last December that summarizes all important aspects: Test Framework Mini Tutorial.

For people building packages that use the new framework, the following new Makefile rules will be of interest:

• make test
  Run all tests recursively from $(TEST_PROGS), abort on first error.
• make test-report
  Run all tests recursively, continue on errors and generate test-report.xml.
• make perf-report
  Run all tests recursively, enable performance tests (this usually takes significantly longer), continue on errors, generate perf-report.xml.
• make full-report
  Run all tests recursively, enable performance tests, enable slow (thorough) tests, continue on errors, generate full-report.xml.
• make check
  Run make test in addition to automake checks.

After GUADEC, we will be looking into getting build machines setup that’ll regularly build GLib, Gtk+ and friends, run the unit testing suites and provide reports online.

For those pondering to write unit tests, but too lazy to look at the tutorial:

• Implementing a test program is very easy, the only things needed are:

    // initialize test program
    gtk_test_init (&argc, &argv);
    // hook up your test functions
    g_test_add_func ("/Simple Test Case", simple_test_case);
    // run tests from the suite
    return g_test_run();

• In most cases, a test function can be as simple as:

    static void
    simple_test_case (void)
    {
      // a suitable test
      g_assert (g_bit_storage (1) == 1);
      // a test with verbose error message
      g_assert_cmpint (g_bit_storage (1), ==, 1);
    }

  Tests that abort, e.g. via g_assert() or g_error(), are registered as failing tests with the framework. Also, the gtester utility used to implement the above Makefile rules will restart a test binary after a test function failed and continue to run remaining tests if g_test_add_func() has been used multiple times.

• Checks in tests can be written with if() and g_error() or exit(1), or simply by using variants of g_assert(). For unit tests in particular, an extended set of assertions has been added, the benefit of using these are the printouts of the involved values when an assertion doesn’t hold:

    g_assert_cmpstr   (stringa, cmpop, stringb);
    g_assert_cmpint   (int64a,  cmpop, int64b);
    g_assert_cmpuint  (uint64a, cmpop, uint64b);
    g_assert_cmphex   (uint64a, cmpop, uint64b);
    g_assert_cmpfloat (doublea, cmpop, doubleb);

  For instance:
  char *string = "foo"; g_assert_cmpstr (string, ==, "bar");
  yields:
  ERROR: assertion failed (string == "bar"): ("foo" == "bar")

• The framework makes it easy to test program output in unit tests:

    static void
    test_program_output (void)
    {
      if (g_test_trap_fork (0, G_TEST_TRAP_SILENCE_STDOUT |
                               G_TEST_TRAP_SILENCE_STDERR))
        {
          g_print ("some stdout text: somagic17\n");
          g_printerr ("some stderr text: semagic43\n");
          exit (0);
        }
      g_test_trap_assert_passed();
      g_test_trap_assert_stdout ("*somagic17*");
      g_test_trap_assert_stderr ("*semagic43*");
    }

• And it is similarly easy to test and verify intentional program abortion:

    static void
    test_fork_fail (void)
    {
      if (g_test_trap_fork (0, G_TEST_TRAP_SILENCE_STDERR))
        {
          g_assert_not_reached();
        }
      g_test_trap_assert_failed();
      g_test_trap_assert_stderr ("*ERROR*test_fork_fail*not*reached*");
    }

• The above and more tests are showcased in GLib: glib/tests/testing.c.

There’s of course lots of room left to improve GLib and Gtk+ unit tests, and also to improve the current framework. For a speculative, non-comprehensive list, here are some ideas from the unit testing section of my personal TODO:

• Introduce 2D marker recognition for graphical unit testing of Gtk+ layouts (prototyped in Rapicorn).
• Provide functionality to determine image similarities to allow for pixel image based unit tests (port this from Rapicorn).
• Implement state dumps to automate result specification and verification in unit tests. (This will allow us to avoid adding lots of abusable testing hooks to our API.)
• Integrate performance statistics (like #354457) and other related information into test reports.
• Publically install a variant of the Makefile.decl file used in Gtk+ to implement the test framework rules and Xvfb swallowing of test programs. This is needed by other projects to run unit tests the way Gtk+ does.
• Implement the unit test ideas that are outlined at the end of this email: Gtk+ unit tests (brainstorming).

Amongst many other things during the Gtk+ Hackfest 2008, it was brought to my attention that Gtk+ maintainership is sometimes perceived as some kind of leet circle, hard to join for newcomers. I can’t really imagine how “hard” it is for newcomers to join Gtk+ maintenance these days. The learning curve probably is steeper now than it was in the last decade, but then, we do have documentation now and didn’t have any back then… ;-)

In any case, I think that definitely none of the Gtk+/GLib core maintainers would want to bar someone else from contributions or helping out with maintenance tasks. So to aid that process, I’ve written down what I keep telling people who approach me about this in person. A lot of it might seem overly obvious to veterans, but for newcomers or contributors looking into extending their activities on the project I hope to provide helpful starting points.

Much of Gtk+ is still maintained as a huge monolithic whole. That is, a very few people are taking care of a wide variety of components. I think ultimately we would all be much better off, if we had more people being responsible for individual components like particular widgets (e.g. GtkNotebook) or widget families (e.g. GtkBox, GtkHBox, GtkVBox). So the following are steps and tasks useful for anyone wanting to get into Gtk+ component maintenance.

First of all, we have the GtkTasks wiki page, a task list with various ways in which people can contribute to the Gtk+ project and start getting involved. In particular for the following positions, no one has signed up so far to volunteer on a regular basis:

• Patch/Bug Monitoring & Filing – collect bugs from mailing lists and IRC to make sure they are being filed.
• FAQ Maintainer – this involves monitoring the mailing list(s), taking CC:-mails from other people on possible FAQ material and regularly updating the FAQ accordingly.
• Build Monitoring – run and maintain Gtk+ tool-chain builds on various platforms.
• Making Releases – we are looking for someone assisting in the release process and to take over release building during some periods in the future.
• Implementing Test Programs – there’s much work needed in our unit test coverage, so we’re always looking for people willing to implement more unit tests.

For general information about the maintenance situation, the State of the Gtk+ Maintenance write-up is still fairly valid. However many good things have been suggested by the community since, such as the GtkTasks page and the Hackfest. At the end, the write-up addresses how occasional contributors or developers at different experience levels can help out the project. For instance with activities such as: Bug triage and verification, Review and clarify documentation, Revision hunting for regressions, Refactor code areas, Work on optimizations, and the list goes on.

If none of this sounds interesting to potential new maintainers, be warned that regular maintenance means having to put up with pretty much all of these at some point. ;-)
However, probably the most straight forward way to take on maintenance for a particular code portion (usually a particular widget) is to start becoming familiar with it and work on improving it right away:

• Cleanup indentation – lots of contributions to Gtk+ in the past have weakened coding style in various areas. In general we use GNU coding style plus a few extra rules similar to the Gimp coding style.
• Perform code cleanups – look for outstanding migrations/refactorings in the code or if the implementation could be simplified/streamlined in areas.
• Check documentation and examples – contributing by improving the existing documentation, testing existing examples and providing new ones is a very straight forward way to learn about a new component. Also, documentation patches are usually easily and quickly approved.
• Provide unit tests – writing new unit tests for existing component APIs is even better than providing documentation examples. You get immediate feedback and they should in general be easy to approve to go into upstream. Also, it is definitely an area where Gtk+ and GLib still need lots of work.
• Review bug reports and patches – go through the Gtk+ bug load of a particular component, see what issues could be closed or need info. Find patches that could be applied or need work and provide/fix patches along the way. Also, feel free to provide review for existing patches where you feel confident to provide reasonable input. For existing maintainers, looking at other people’s review is one of the best ways to figure if another person is up to the task of taking over component maintenance.
• Actively nag existing maintainers or people with SVN accounts for trivial patches (like Cody and Mathias who signed up for Patch testing & committing) to review, approve and apply your changes.
• Participate in the project forums like gtk-devel-list and #gtk+ (needed to nag people from the core team and other developers), and read and reply in other related mailing lists.

The first few points actually mean working with the code by providing patches, and filing new bug reports with those patches attached. While this may at first increase our bug load, if someone shows sincere interest in taking over component maintenance, sticks to the coding style and provides useful patches, there’s nothing stopping us from granting new SVN accounts so contributors can commit their own patches after approval.

Finally, the project is welcoming every new contributor. But be reminded that no approval is needed from anyone to start your work. In actuality, asking for it will most probably get you a reserved or negative answer, because improving the project is all about working on it, not making or requesting promises. So, everyone is invited to read through the task lists/descriptions and get their hands dirty immediately. A fair bit of self incentive is needed to take on maintenance of a component anyway, so you’ll have to get yourself motivated on your own there won’t be anyone else doing it for you.

There have been several requests about hosting Gtk+ (and GLib) as a Git repository recently and since that topic has come up more and more often, I meant to write about this for quite some time.

Let’s first take a look at the actual motivation for such a move. There are a good number of advantages we would get out of using Git as a source code repository for Gtk+ and GLib:

• We can work offline with the Gtk+ history and source code.
• We can work much faster on Gtk+ even when online with tools such as git-log and git-blame.
• It will be much easier for people to branch off and do development on their own local branches for a while, exchange their code easily with pulling branches between private repositories, etc. I.e. get all the advantages of a truly distributed versioning system.
• With Git it’s much easier to carry along big Gtk+ changes including history by using cherry picking and (interactive) rebasing.
• We can make proper public backups of the source code repositories. This ought to be possible already via svnsync, but isn’t for svn.gnome.org because we run an svn 1.3.x server instead of an svn 1.4.x server that is required by svnsync. (Yes, this issue has been raised with the sysadmins already.)

A quick poll on IRC has shown that all affected core maintainers are pretty much favoring Git over SVN as a source code repository for GLib/Gtk+.

However, here are the points to be considered for not moving the repositories to Git (just yet):

• Complexity; Git is often perceived to be harder to use than SVN, there are several attempts to mitigate this problem though: Easy Git Giggle yyhelp.
  With some of the above, Git is as easy to use as SVN is, so Git complexity doesn’t need to be holding anyone off at this point.
• Git may be stable and mature these days, but git-svn is not there yet. It is generally good enough to create local Git mirrors of SVN repositories and work from those to have most of the Git convenience on top of an SVN project.
  But git-svn has seen structural changes recently, quite some rewriting and bug fixing that indicate it’s still too much in flux for the one-and-only SVN->Git migration for projects at the scale of Gtk+. This is not meant as criticism on git-svn, fairly the opposite actually. It’s good to see such an important component be alive and vivid. All issues I’ve raised with the maintainer so far have been addressed, but it seems advisable to wait for some stabilization before trusting all the Gtk+ history to it.
• Gitweb interfaces already exist for GLib/Gtk+ mirrors, for example on testbit: Testbit Git Browser.
  These can be used for cloning which is much faster than a full git-svn import. Alternatively, shallow git-svn imports can be created like this:

    git-svn clone -T trunk -b branches -t tags -r 19001 svn://svn.gnome.org/svn/gtk+
  

  This will create a repository with a mere ~1000 revisions, including all changes since we branched for 2.12. We’re using such a shallow repository for faster cloning of our GSEAL() development at Imendio: view gtk+.git.

• In summer 2006, we’ve had the first test migration of all of GNOME CVS to SVN, in December 2006 we’ve had the final migration. During that period, the Beast project stayed migrated to SVN to work out the quirks from the CVS->SVN migration before we migrate all other GNOME modules and have to fix up everyones converted modules. There were quite some issues that needed fixing after the initial test migration and in the end we had to rebuild the Beast development history from pieces. Preventing the other GNOME modules from such hassle was the entire point in migrating Beast early on, so I’m not complaining.
  However, given the size and importance of GLib/Gtk+, the development history of those projects shouldn’t be put at a similar risk. That is, GLib/Gtk+ shouldn’t be pioneering our next source repository migration, let some other small project do this and work out the quirks first.
• git-svn already provides a good part of the Git advantages to developers while Gtk+ stays hosted in SVN. Albeit due to mismatching hashes, syncing branches between distinct git-svn checkouts of different people is tedious. Setting up an “official” git-svn mirror on git.gtk.org could probably help here. Also, to ease integration, jhbuild could be extended to use git-svn instead of svn commandos to update SVN modules, if the current module has a .git/ subdirectory instead of a .svn/ subdirectory.

The bottom line is, there’s a good number of advantages that Git already can (or could) provide for our development even without migrating the repositories to Git right away. When exactly will be a good point for migrating GLib/Gtk+ and possibly other GNOME modules might not be an easy call, but right now is definitely too early.

I finally got around to fix a long-standing and tricky bug report: Bug 64764 - Class initialization isn't thread safe.
Thread safety problems in class initializers and _get_type() functions caused nasty problems in other components that depend on parallel type creation, in particular GStreamer (Dependency Graph for bug 64764). With both being fixed now, testing feedback about GType/GObject threading problems using GLib trunk is appreciated.

I’m currently in the process of building some general purpose infrastructure for Rapicorn and because a good portion of the C++ utilities used in Beast and Rapicorn are factored out into an extra library called Birnet. I’m running into similar complexity issues here that we encountered with libglib vs. libgobject vs. libgdk already. I.e. lower level modules cannot reference or use mechanisms provided at a higher level (for instance deriving GHashTable from GObject). This becomes even worse with C++ which lacks support for partial classes (a class definition spanning multiple source files, like implementing gtk_widget_foo() outside of libgtk).

One remedy would be to actually merge libbirnet and librapicorn into a single project and let Beast depend on librapicorn. However, that’s essentially like we started out with GLib and Gtk+ in late 1996 when it got split out of the Gimp package. In June 1998 we split off libglib into a separate package for programs needing C utilities but not libgtk. So now, I’m reconsidering this separation after almost a decade. Reflecting on the move is probably not a bad idea before attempting the opposite with Rapicorn:

How many programs are actually linking against libglib but not libgdk, and is there any real benefit from separation?
If your machine has libglib and a few spare minutes, you can use this command line to find out on your own:

	find `echo $PATH | sed 's/:/ /g'` -type f -perm +111 | xargs ldd 2>/dev/null |
	  awk '/\<libglib-2\>/{ lg+=1 } /\<libgobject\>/{ lo+=1 } /\<libgdk_pixbuf\>/{ lp+=1 }
               END { print "glib=" lg " gobject=" lo " gdk=" lp ;
                     print "glib-only:    " 100*(lg-lo)/lg "%";
                     print "gobject-only: " 100*(lo-lp)/lg "%";
                     print "gdk-lot:      " 100*lp/lg "%"; }'

Here are results from a couple machines I had quick access to:

	Site                       Gdk  GObject     GLib  #Apps
	gimp.org:                50.0%    12.5%    37.5%     16
	gtk.org:                 47.1%    17.6%    35.3%     17
	developer.gnome.org:     19.5%    53.7%    26.8%     41
	my server (sarge):       66.7%    14.1%    19.3%    192
	my laptop (etch):        72.7%    14.8%    12.4%    209
	my desktop (feisty):     72.2%    17.1%    10.7%    252
	64bit KDE desktop (sid): 53.0%    16.7%    30.2%    338

That is, the servers have a quite limited set of GUI applications installed, but across a full fledged desktop the vast majority of applications is linked against libgdk anyway. I found Stefan Westerfelds Amd64 KDE desktop particularly interesting: It actually has the most applications linking against GLib, prolly because it has Mono and libQtCore.so installed which both link against GLib these days. Does anyone actually have a system with libglib installed but not libgdk?

Next, let’s take a look at the actual “savings”:

	ls -l libglib-2.0.so.0.1200.4 libgmodule-2.0.so.0.1200.4 \
              libgobject-2.0.so.0.1200.4 libgthread-2.0.so.0.1200.4
	-rw-r--r-- 1 root root 596608 2006-11-16 10:26 libglib-2.0.so.0.1200.4
	-rw-r--r-- 1 root root   9784 2006-11-16 10:26 libgmodule-2.0.so.0.1200.4
	-rw-r--r-- 1 root root 237156 2006-11-16 10:26 libgobject-2.0.so.0.1200.4
	-rw-r--r-- 1 root root  14028 2006-11-16 10:26 libgthread-2.0.so.0.1200.4
	=====================================
	                       857576 = 837KB

	ls -l libatk-1.0.so.0.1214.0 libcairo.so.2.9.2 libpango-1.0.so.0.1400.8 \
	      libgtk-x11-2.0.so.0.800.20 libgdk_pixbuf-2.0.so.0.800.20 \
              libgdk-x11-2.0.so.0.800.20
	-rw-r--r-- 1 root root  102504 2007-03-14 13:44 libatk-1.0.so.0.1214.0
	-rw-r--r-- 1 root root  395836 2006-10-20 07:44 libcairo.so.2.9.2
	-rw-r--r-- 1 root root  235464 2007-01-14 22:27 libpango-1.0.so.0.1400.8
	-rw-r--r-- 1 root root   88604 2007-03-04 22:21 libgdk_pixbuf-2.0.so.0.800.20
	-rw-r--r-- 1 root root  528580 2007-03-04 22:21 libgdk-x11-2.0.so.0.800.20
	-rw-r--r-- 1 root root 3043524 2007-03-04 22:21 libgtk-x11-2.0.so.0.800.20
	=======================================
	                       4394512 = 4292KB

GtkGdkPangoAtkCairoGlib / GLib ratio: (4394512 + 857576) / 857576 = 6.12.
So the “savings” turn out to be splitting off 5/6th of the GUI stack size. At best, a GLib-only program is saving 4.2MB that way. But then again, those are likely to be in memory already anyway. And most definitely, they reside in a library, available as const .text on the harddisk. To put that into perspective: We’re talking about only a handful megabytes here. In physical size actually less than a hires desktop background image, smaller than some icon themes, or roughly 10% of a full linux-2.6.8 kernel + modules binary footprint (41MB). Also symbol resolution performance doesn’t necessarily improve through splits, Ulrich Drepper has a few words on multiple DSOs during deployment.

Given that selecting Gtk+ subcomponents to allow selective shrinkage for size constrained embedded devices via configure.in options is on our TODO list anyway; for this set of libraries: Size is not worth a split!

Of course, other aspects not reflected in library size weigh in much more strongly. Such as the ability to build different user communities around projects (e.g. for Cairo vs. Glib) or the reduced dependency chain (sshfs and Qt wouldn’t depend on GLib if it drew in Pango or Gtk+).
So while for the vast majority of applications the GLib split off doesn’t make a huge difference technically, software evolution arguments make all the difference, definitely justifying a separation. The GLib internal splits are definitely not worth the hassle and conceptual boundaries though, if we ever break binary compatibility again, merging the GLib internal libraries and respective pkg-config packages would be a good idea.

Back to my original problem… I’m now pretty convinced that merging Rapicorn with the Birnet utility library will most likely not pose a problem anytime soon. And Qt 4 kindly demonstrates that splits can also be carried out successfully later on during the game.

So i sat down to implement a slice address and size validator to catch the most frequent GSlice misuses. The debugging validator consists of a flat fixed-size hashing tree with fairly big prime-sized nodes and binary searchable arrays to manage collisions in the hash buckets. This structure avoids circular dependencies with GSlice (like GHashTable and GTree would have it) and is still reasonably easy to implement. The hashing makes use of the approximately random distribution of allocator block addresses in the lower bits but fairly sequential distribution in the higher bits, so the structures are reasonably space efficient and can adapt in size incrementally with every new megabyte handed out by malloc(3)/memalign(3). This also keeps hash collisions on an acceptably low average for allocations within a 4GB address range. Using a single global structure keeps GSlice from scaling well across multiple threads, but i think that’s fair enough for a debugging mode validator. Quickly after running an early version of this validator, the cause of the above Beast bug was found and could be fixed.

	$ G_SLICE=debug-blocks testcase
	GSlice: MemChecker: attempt to release block with invalid size: 0x80527e0 size=8 invalid-size=12
	Aborted

The faulty code was mixing up GSList and GList nodes, which couldn’t be caught by the compiler because the use of object data and timeout data forces void* casts and gives up type safety:

	void *slist = g_slist_alloc(); // void* gives up type-safety
	g_list_free (slist);           // corruption: sizeof (GSList) != sizeof (GList)

At this point, the validator is in GLib SVN and can be enabled with G_SLICE=debug-blocks.

Another issue that some people might run into with the recent GSlice code is:

	***MEMORY-WARNING***: GSlice: g_thread_init() must be called before all
	other GLib functions; memory corruption due to late invocation of
	g_thread_init() has been detected; this program is likely to crash,
	leak or unexpectedly abort soon...

The correct workaround for that is to add an early g_thread_init() call to your program. More details can be found in the gtk-devel email: bugs regarding late g_thread_init() calls.

	G_SLICE=always-malloc G_DEBUG=gc-friendly valgrind --leak-check=full mybuggyprogram

Or, in combination with MALLOC_CHECK_, it can be used to catch and debug invalid allocation/release calls:

	G_SLICE=always-malloc G_DEBUG=gc-friendly MALLOC_CHECK_=2 gdb mybuggyprogram

And, the hopefully last incarnation of the atomic ref-counting patch for GObject i blogged about earlier finally made it into CVS. With atomic ref-counting, GObject users still have to make sure they don’t modify or read-out the same object concurrently from different threads (such as using g_object_set_qdata() on it or using object methods), they may however invoke g_object_ref() and g_object_unref() concurrently without having to ensure mutual exclusion, the same goes for GParamSpec and GClosure.