Stuff Michael Meeks is doing

• Up early, dealt with the babes while J. slept. Off to NCC, Tony spoke. Home for a fine roast lunch with R & I.
• Slugged around the house, then out for an Alleluja party (the antidote to Halloween) at All Saints - lots of small people, dressing up as animals, doing all manner of simple games. Out to the pub with DT & Robert.
• DT, Zoe & Patrick over in the evening, crumpets and tea by the fire; talked until late with I&R - sorry to see them so infrequently; bed late.

• Breakfast; off to Sizewell beach to play with the babes - lovely sunny day. Spent lots of time moving gravel into banks to funnel the incoming waves into a deep channel up the beach - fun but exhausting.
• Back via Aldeburgh for a very pleasant fish & chip lunch. Packed, and drove home - unpacked machines. Painted the external doors with a second coat of superdeck - they still don't look an even color (most odd), but at least somewhat better. Perhaps my rolling is too aggressive.
• Robert & Ilona arrived late, helped put babes to bed, and had dinner with them. Up late talking over the world and its troubles.

• Boggled at Oracle's amended complaint against Google. Havn't investigated the verbatim code copying claims in detail - which are of course fair game for copyright enforcement - but what interests me -far- more are the API copyright claims:
• As I wrote before, I believe that Java has rather weak patent protection, not being particularly innovative or novel really - rather a collection of existing techniques packed together into one (useful) package. With a load of design artefacts in the API that can be claimed as in some sense 'novel' though of little intrinsic technical merit. I wrote a long screed on the background to this: Why Oracle's Java Copyrights Might Matter in August. As I also wrote before: I am not a Lawyer, and have no very deep understanding of the Java situation.
• As if to offer further proof, Oracle now have amended their complaint to include this:

  Android includes infringing class libraries and documentation. Approximately one third of Android's Application Programmer Interface (API) packages ... are derivative of Oracle America's copyrighted Java API packages ... . The infringed elements of Oracle America's copyrighted work include Java method and class names, definitions, organization, and parameters; the structure, organization and content of Java class libraries; ...

  This seems (to me) to be an extraordinary claim.
• As previously discussed, if the claim is that APIs can be copyrighted, the whole world of interoperability will take a staggering step backwards into the dark ages of walled gardens, and pointless differentiation. Perhaps Oracle itself has never, and will never interoperate with others by providing drop-in replacements of a competitor's API - but it seems highly unlikely - though whether estoppal will be kind to them in future in this area is unclear.
• To me this smacks of desperation and highlights the weakness of any patent protection around Java; after all, if there was strong patent case, why include a dangerous chestnut like this? Particularly one that may provide a deeply negative precedent for the wider industry: unfortunate. Hopefully in terms of this case, on this point - that will provide a set of useful allies for Google in the form of amicus briefings on the topic.
• Some questions are left un-answered: how far does this supposed copyright protection go? as an example if you implement a Java Listener interface - you have to implement half a dozen methods with identical method names, identical parameters, definitions etc. to those in the Java listener interface - in your own code, eg. the WindowListener interface. Something like this (and my Java is rusty):
  

  class Dangerous implements WindowListener {
          public void windowClosing(WindowEvent e) { do something; }
          public void windowClosed(WindowEvent e) {}
          public void windowOpened(WindowEvent e) {}
          public void windowIconified(WindowEvent e) {}
          public void windowDeiconified(WindowEvent e) {}
          public void windowActivated(WindowEvent e) {}
          public void windowDeactivated(WindowEvent e) {}
          public void windowGainedFocus(WindowEvent e) {}
          public void windowLostFocus(WindowEvent e) {}
          public void windowStateChanged(WindowEvent e) {}
  ...
  }

  Does 'copying' that listener signature give Oracle a similar copyright claim on your code? surely not. What if you implement dozens of listener interfaces all with methods matching Oracle's (supposedly copyright) API - does that give Oracle an ownership right to parts of your code? again it seems incredible. What if you used one of the hundreds of books that comes with sample code that looks just like the above to implement your code ? do Oracle own you ? What if you use the same argument names as the sample in the book ? What if you take the constructor example, and turn those argument names into member variables to store the data in (as people do) ? Then again, you can see this sort of code licensed under a BSD license from Oracle themselves; how much of the supposed API is intrinsic to existing BSD licensed examples ?
• I would point out, again, that the correct way to go about publishing standards (whether openly developed, as they should be, or not) is to submit them to ISO (and/or first ECMA) such that there can be no copyright issues, and that the language / run-time & so on can be implemented by all without fear of bogus API copyright attacks. It is also nice if there is a patent covenant for all implementers from the original creator.

• Up early, dug through Oracle's amended complaint vs. Google, sad; wrote a bit about it. Very encouraged to see Kay's nice eight second full-boot chart using systemd.
• Dug into a strange dependency problem - that magically compiles files not mentioned in the makefile in some cases. Lunch. Started prodding at the migration code, some most interesting code It is amazing what we do in the SEGV handler - taking locks, writing the document out etc. sadly of course a SEGV can happen at an inconvenient place - such that even calling pthread_cond_signal or somesuch can cause deadlock if the condition is locked - as used to happen to Mono sometimes before it was fixed some years ago.
• Dug through the wife's mind wrt. charity law, good to have an on-hand expert.

• Boiled eggs and soldiers for breakfast with the babes. To work, plugged away at the E-mail and patches much of the morning. Avoiding expense reports. Lunch.
• Got 'experimental' option patch done, and set-off a new compile to test it; several hours even with the cluster. One of the most curious things about Novell is that they pay their most junior clerical administrators a programmer's salary; or - put another way, they encourage us to spend a solid three hour block of time every now a few times per year simply sorting, collating, and entering expenses into a lameish web tool. Of course, the ideal is to dis-aggregate that so it is not noticed, but somehow - on arriving back from travel, the pressure to work through not only the mail backlog from the journey, but also the exciting / new tasks from the trip is too intense, and the tiresome admin inevitably gets dropped.
• Build finally completed, pushed the new option, more mail, knocked off for dinner. Prodded at UK law in the evening a little, whole schedule knocked off track by administration.

• Up lateish, prodded the mail mound; got my icecream USB bootable build-slaves building (improved studio image to follow), boggled at Twitter's (huge, graphical) "Twitter is over capacity." whale artwork; not that I ever use the thing, so its no loss.
• Another really nice presentation in the openSUSE conference / LibreOffice track was from Jonas - who has done some great work making the OO.o formula editor properly interactive. We are merging this into LibreOffice 3.3 as the first 'Experimental' feature: ie. features you can enable at your own risk, to try out cool new things:
  
  The great news about this experimental feature is that, (of course) while insanely cool, it is not perfect yet. There are a number of easy hacks available just for this piece: starmath is quite a nice, self contained piece to get started in. Other sexy things in Jonas' work (apart from the code) was a keybinding to dump the scene graph to a file, combined with a watcher that throws up a GraphVis graph of the structures: very sweet. I wonder if there is scope for integrating GraphVis into LibreOffice in some useful / WYSIWYG way.
• Dug into some evil visibility annotation problem relating to multiply defined abstract derived class type information afflicting the split out MS word filters.
• Drove to Bruce & Anne's, and re-assembled the compile cluster quickly - cheap netbooks make quite nice, small network-attached CPUs. Discovered that Gnome Shell has a reall nice set of tagged easy hack tasks.
• Finally noticed that Florian's LibreOffice slides are up. Reviewed and merged some more patches, listened in for part of the Project Harmony call.
• First public LibreOffice steering call in the evening, worked late on this and that.

• Out for breakfast at Faith's home - great to see her, Jill, Laura and the collective menagerie of children; home to work.
• Amused to see the Canonical noise around their new desktop shell. As Dave says writing a functional shell, particularly with a team with a steep technical curve to climb, and few spare resources. Perhaps it will succeed in the long run (where by 'succeed' I mean create a really viable duplication of Gnome Shell which can run and run, dividing developers and users), and perhaps not. Intel also initially made a ton of mistakes in Moblin around copyright assignment and closed development which breed duplication etc. though they seem to be past much of that now. Arguably Novell's development of compiz itself was extremely retarded wrt. open-ness, inclusion and community nous. Making the same silly mistakes left and right is nothing new. The sad thing in the bigger picture, is to see the community of companies grow (and fail) faster than the lessons / experience of past failures percolate through to their leadership. In my experience most people totally miss the most difficult piece of software engineering: which sounds like it should be software - but is really about people - and more importantly - them working together collaboratively in a constructive, friendly, and incremental fashion. Obviously collaboration is not always possible, but in the long run - there is simply no other viable approach than to work well with others, (in my not so humble experience).
• Prodded mail, really encouraged to see VOS (a library deprecated for a ~decade) finally expunged from LibreOffice by a volunteer - Norbert Thiebaud - a hero of our time; lots of code review, chunks of testing, and a cleaner, smaller, more consistent code-base. Reviewed and merged a nice patch to use zenity during the build for systray / progress feedback.
• Amused to see Oracle's Community Council show its true colors; with everyone except Oracle guys leaving - apparently this non-entity (the council), with no actual legal existence needs change. Sadly its bylaws don't allow for Oracle unilaterally deciding that it doesn't want to hear the elected community representatives say - which is no doubt a bug. To try to portray this as a conflict of interest of the community representatives, there being no conflict of interest for the Oracle guys is a joke. Why should a divergence from Oracle's view be a conflict of interest with the community ? I guess the tactic here is to push the lunacy one level up, and claim that because Oracle owns the trademark (and refuses to donate it to the community) that per-se the 'OpenOffice.org Community' is owned by them - but this seems horribly circular. Some people have not resigned formally until 3.3 is out, since that was their stated intention I think.
• Meanwhile Miklos posted his slides from the LibreOffice track at the openSUSE conference. His talk "GSoC project: Improving RTF Export..." is available as PDF. goes over the merger of the RTF export filter into the new combined MS export filter code. This saved lots of lines of code, removed tons of duplication, and resolved a lot of bugs; a couple of larger scale links here:
  
  I am always amused to hear the insinuations of how 'un-professional' and 'low-quality' our code contributions are, but this "Anything in a document after a formula is lost when saving as RTF" bug, present in OpenOffice.org (and fixed in Miklos' work - as well as making these formulae editable by the way) takes the biscuit, though there are a large number of other great fixes included:
  
  See the talk for another dozen examples of key improvements.
• More mail, Clarity, Lunch, fixed make_installer breakage caused by including Digest::MD5 in solenv by reverting it; hmm. Debugged an f-spot problem for Stephen. Worked on mail queue, added some more easy hacks.
• Lydia and Andy over in the evening for dinner, sat by the fire for a while talking, bed late.

• Up late, plugged away at mail, call with JP, then Charles, back to JP. Prodded at patches.
• Great to see Cedric's write-up on the LibreOffice contribution statistics for the first weeks. Here is his graph of contributors:
  • Clearly lots of these contributors are pushing fixes and cleanups, not huge features (yet), indeed some of the changes are quite small, but then, of course - you would be mad to work on a huge feature without knowing if you can first get a small fix included; so this is expected. Also by cleaning the code up, we enable yet more contributors in a virtuous circle. Though the Oracle contribution seems to shrink, this is most likely just an artifact of weekly reporting, and the place in the stabilisation cycle - I don't believe they have reduced their investment here.
  • Week 38 is when LibreOffice was created - what changed then ? two similar things: Vendor neutrality, and dropping the requirement for a Copyright Assignment - otherwise that is about it. This combination adds up to true open-ness.
  • The graph does not show that we had sixty-five 'new' contributors (ie. not including existing RedHat Novell, Debian, Oracle etc. hackers), but that number becomes stale daily as new people get stuck in.
  • Really exciting times in LibreOffice, see if there is an Easy Hack you can get your teeth into today.
• It is just so wonderful to be home with the lovely wife after a week away, complete again. Reviewed and merged a nice pam removal patch, build server PSU died taking out the whole house' power momentarily - lovely smell of vaporised electrolyte; thank God for metal PSU enclosures.
• Discovered ace_dent - oh he of the LibreOffice icon compression work - is in fact supervising a PhD student on the reprap project - which is awesomeness in a nutshell. Failed to get all my administration and E-mailing pieces under control, but (mostly) dealt with the urgent patch reviewing task.

Reprap ?

The reprap project provides the designs and drawings necessary to make a self replicating, plastic extrusion 3D printer. There is only one hitch: self replication; you need a plastic printer in order to make one. Failing that you need to bootstrap your printing world by building a 'repstrap' - something that can (just about) print the pieces.

An ad-hoc repstrap

I have knocked together such a thing from spare bits of floating timber (mostly MDF and ply-wood), which has some unpleasant calibration problems: related to thermal issues, and a lack of an adjustable, (or even better heated) bed. Nevertheless it can be persuaded to print useful things. Indeed - arguably it is easier to cut a nice, rigid sheet of MDF, than to create isometric triangles made of studding with complicated plastic verticees. Anyhow - the beastie re-uses almost all the parts: steppers, controllers, belts etc. from an existing reprap - so you can re-use them in the final product (in theory). It also requires only simple hand-tools (except for the extrusion nozzle, and knurled driver which is a pain generally). It looks a little bit like this:

Notice the horrific warping problems that come from poor print head - bed spacing calibration, and the lack of a heated bed. Still, this is one of the largest, and most problematic prints necessary to bootstrap. I have the vague intention to create useful drawings, and instructions to replicate/iterate it for others with low-precision tooling - but the joys of LibreOffice are ship-wrecking this aspiration currently. Please also note that the all-important safety goggles are just out of frame somewhere.

Making large z axis gears

One of the most fun problems of repstrap construction are the gears. After several attempts - to whittle gears from acrylic rod, to make them from MDF, and so on I hit on a near perfect solution, reproduced here for your gear-constructing pleasure.

It turns out that simple 30mm panel-pins fit rather snugly into the required timing belt, creating a nice firm fixing. So - this makes the construction of the large gears a matter of rough cutting vague disk shapes, (or better using a nice large circular cutter) four disks from thin ply-wood, then using the stencil linked below - marking out the pin locations, drilling them through (as vertically as possible) with a 0.5mm drill, and then simply pressing the pins through. This and a pair of nuts in the centre yields a beautiful, functional z drive, as modelled by my youngest, as she concentrates on reprap assembly thus:

Stencils for gears

It turns out to be remarkably hard to draw gears, or lay them out manually by measurement. Luckily most people have a phenomenally accurate ink-jet printer. Thus simply print out this file: gears.odg - clearly using LibreOffice if you don't have it. Then use a bradawl to make holes to enlarge with a drill in the right positions.

Making stepper gears

These smaller gears present a real problem for ply-wood - which would just crack and split, and not provide enough purchase on the shaft of the stepper. After some initial semi-functional attempts with some aluminium, I discovered that the perfect material for small pin gears is near to hand, and luckily is stocked in your kitchen, it looks like this:

Naturally, it is only a matter of explaining to a suitably patient wife, her desparate need of a new chopping board, and the invaluable nylon is procured. To satisfy yourself that it is indeed a wonder material, try driving a panel pin into it 5mm from the edge, and then breaking it out. Attack the chopping block with a junior hack-saw or some such device, to make an octagonal piece of suitable diameter. Again use the bradawl, stencil, and 0.5mm drill as above. Drill a larger central hole, which should be of a slightly smaller diameter to the drill shaft (which should have flats filed onto it). Simply 'drift' (ie. wallop with a hammer) the nylon onto the shaft for a fool-proof fix. Eventually you end up with something beautiful like this:

Why bother ?

Hopefully, after much more wood hackery, adjustment, suffering, remedial medical treatment and so-on you can finally produce a working 3D printer, and hence something actually worthwhile. While this is only a prototype - and died on the first attempt of a lawyer to bend it, the full (more solid) version has resisted the attempts of a long line of hackers in their eagerness to break the end off; enjoy:

Of course, you really need to use RepSnapper on openSUSE for controlling that sort of thing.

• Up lateish, met Egbert - good to catch up with him; breakfast with Fridrich and Thorsten - bid 'bye to them. Off to the office, prayed, enjoyed a rather excellent sermon from Park Street: Don't be afraid: just believe. Met a nice chap: Doc Hodges in passing.
• Out to a cafe for the afternoon with Garret - great to catch up - a pleasure sadly deferred for some years. Off to the airport. Flight; tried to get through the Economist after a frantic week. Decided (finally) to post my reprap pictures

• Woken by a text message, having somehow missed the alarm. Rapid dressing, off on a walking tour of Nurnberg - lead by Stefan Behlert - who turned out to be an expert tour-meister. Enjoyed the sights of this beautiful inner city.
• Sausages in a bun, then off to find some steak. Most impressed to discover Miklos had hacked up a German comment detector overnight, using the OO.o built-in language-guessing code. Hopefully it is not be generally useful for other projects, but invaluable if you have an in-code language problem.
• Off to the office to hack with the lads; worked over some Calc issues with Kohei again, merged some code; dragged from the hacking for dinner (Japanese) with the remaining team members, and back for drinks before sleep.

• Off to the conference venue in a taxi with Mathias & Alan. Enjoyed Gerald's talk, then on to Florian's opening LibreOffice talk, brainstorming / BOF on features with Thorsten.
• Fridrich's (well attended) Win32 / build-service talk, back to the excellent talks from Miklos & Jonas - most impressed with their progress, and (reading the code later) the quality of the work.
• Caught up with journalist friends before heading back to the office, by bus with the lads - out to eat - at the same restaurant with the team, stayed up rather late talking of this & that.

• Up, breakfast with the lads; out to the office, chewed mail a bit, chat with Dave. Off to the conference centre, spent lots of time talking to people, with a snatched lunch in between.
• On to Lennart & Kay's systemd BOF - exciting to see so many interested people, new features in action; and the calibre of the attendees.
• Back for the evening party; sadly missed the food, but spent lots of time talking to the great people we have around the place. Back to some typing very late in the evening, while people hacked at slide-ware for the next day.

• Up, breakfast with the lads; off to the conference with Lubos. Caught up with Andrew, gave him the stage for the Smeegol talk ( since he does all the work it seems only fair ), nice talk.
• Interview with a pair of friendly journalists, quested for Lubos, back to the office, and out for Thai food with the team. Much discussion, charts, movie clips, discussion & fun. Pizza and spreadsheet internals with Kohei (to aid digestion).
• Lennart, Kay, Alex & Federica arrived, caught up with them, and the latest, sexiest bits in systemd. Bed late.

• Up truly early - 4am, staggered to the coach, wandered through check-in and security; used beautiful Android ethernet / 3G tethering to be productive at the airport, reviewing patches, setup another couple of git accounts, and mail.
• Finally saw that the great honour of a Plumbers Conf closing keynote got announced: "The wonderful world of LibreOffice" - with particular reference to plumbers: ie. a bus-load of low-level grot, tooling and fun. Of course, what with the Plumbers conf being insanely cool, it is sold out - so prepare to sell your home, to be able to afford your tickets on the blackmarket in Boston.
• Caught the plane, seated next to a friendly journalist; met yet another as I got off - great guys going to the openSUSE conference, shared a Taxi to the hotel.
• Off to the office, wandered the halls harassing the innocent who failed to escape in time; we have some great guys. Plugged away at some numerical tasks, and then out for lunch with the Czech lads.
• Back, roadmap plotting, misc. meetings, more and more Novell LibreOffice team people arrived, out for dinner in a very cosy place for a fine German meal, bed late after a re-assuring call with Andrew.

• Prodded mail, interested to read Bradley's heated piece on Open Core, also known as 'The valuable[sic] MySQL model'. In my view this model leads, almost inevitably, to community dysfunction, and total corporate dominance of whatever piece it is applied to; perhaps I am wrong, time will tell.
• Booked coach for 4:40am tomorrow - ouch, still really looking forward to the OpenSUSE conference, meeting up with the attendees, and the LibreOffice guys to plan the future.
• Lunch with Lottie, Claire popped over after that. Back to a call with Kendy. Spent a good while reviewing another huge patch set removing bus-loads of non-compiled code from LibreOffice - heavy reading indeed. Chat with Dave Neary to brighten the tedium.
• Annoyed by this somewhat false dichotomy being presented routinely these days on the lists, to try to drive a difference between the LibreOffice community and OpenOffice.org. Now, it seems, there is (sadly) a big barrier of rejection between Oracle and LibreOffice - they seemingly don't want to join yet; but OpenOffice.org should not equal Oracle, and the charge of conflicts of interest cuts two ways. To distort Simon Phipps' quote a little: "If the company sponsor stands still and the community moves on, who has the conflict of interest ?". It is amazing (to me) that some people know the will of the community better than every single elected representative from outside Oracle that is on the council.
• Booked my flight to the MeeGo conference and remembered to register on-line as well. Packed everything into one small bag, and bed earlyish.

• Up earlyish, off to NCC - visiting speaker spoke on troubles, peace, and joy. Home to prep for lunch.
• Chris, Joy & daughters over for lunch from Oak Hill - lovely to see them, fantastic to have sorted relatives. Enjoyed their company for some hours, first fire of the year - starting to get rid of the huge stack of wood left over from the building work.
• Sogged in front of some children's DVD with the babes, light tea to make up for the mammouth lunch; bed early.

• Up early, dropped remarkably reasonable babes off with the Brighty's, and off for another day session again with Steve Hall.
• Studied the Five Circle theory: Physical, Emotional, Volitional, Rational, Spiritual - at some length, most interesting. Ellis' A-B-C-(D-E-F) theory, Did a little self-awareness work - scary enough.
• Back to collect babes, read stories, dinner, bath, bed, lay around talking until late.

• Up early, prodded mail, ran tests builds, did some product analysis. Knocked out my Mannah House homework: "List some of the key elements / concepts of, and differences between Psychoanalytical counselling, Person Centred counselling, and Rational Emotive Therapy." and mailed it off. Lunch.
• Fixed lpsolve dependency problem; $(IFEXIST) doesn't work on directories (it turns out). Analysed the output from another of Julian's wizard valgrind tools run on LibreOffice. Added another easy hack or two. Tried to dig through the torrent of good mail & patches flowing into the inbox.
• Dinner with Mary, out with J. to counselling evening with a new lady - Lynn; amusing evening about pratical counselling, communication, etc.

• Prodded at mail; attacked the build again, made the LibreOffice build script give helpful messages on build failure, and configure warn about expected problems with non-default --with-distro flags, binned obsolete patches.
• Lunch, team meeting. Prodded at systemd a little in the openSUSE 11.4 milestone two into the boot sequence, with a nice wiki page explaining any openSUSE issues. Eventually gave up, and installed it in KVM to avoid lots of re-booting - sadly to save space had to remove my opensolaris image: RIP.
• Conf call, dinner during it; put babes to bed while J. counselled, more work, call with the parents - just got home. Worked late merging patches & fixing associated bits.

• Up early, off into Cambridge with Mike; TEAM training, all day - lots of interesting people, lots of good talks on Galatians, Evangelicalism, and then small-group practice in creating and leading a short Bible study; fun.
• Back for dinner with the babes, played with them, stories, put them to bed, out to Cell group.
• Interested to read Matthew Garret's take on Unicode 6.0, which is of course an encouraging release. What would seems sad to me though, is an outcome where a picture of men, or women holding hands was to come to mean a symbol of a homosexual relationship; straight men are often isolated enough without sterilising a sphere of legitimate and in many cultures (Arab, African etc.) normal form of innocent intimacy.

• Up early; mail poking, interested by the news of IBM joining OpenJDK the real un-answered question is: what sweetheart license terms have IBM extracted from Oracle, or will they be distributing the code under the GPLv2/classpath like everyone else ?
• Prodded at dependency errors in the build; made the unix quick-starter use libpng - since we require it anyway, and made it conditional (only for distro builds: where we have a known libpng we can link to). Made broffice branding disabled by default for non-LibreOffice distros, to avoid surprises.
• Fiddled with the website, spaetz moved the LibreOffice Easy Hacks page to the nicer DocumentFoundation (media) wiki. Prodded the scp2 dependency problems.
• Dinner, put babes to bed, chat with Patrick & Tor. Chat with Mike. Pushed a fix to make scp2 (the installation magic) re-compile if you re-configure LibreOffice with different options. Merged a couple of community patches.

• Up, read mail, re-factored splash, about and shell branding in LibreOffice to allow a single build including BrOffice branding in Brazil; dunged out lots of old / odd code and artwork, and eventually pushed it.
• Misc weekly status admin, reviewed LXF column, more patch review and bug chasing.

• Up early; dealt with babes while J. slept - dropped them at All Saints for the harvest festival. Went on to NCC for Christine's daughter Hannah's dedication - fine service. Back to All Saints to crash the rather fine shared lunch with the family.
• Home, moved shelving up-stairs, and J. got stuck into tidying my office - always a nerve-wracking time. Good to finally get all those useful reference books back into view, and on some shelving.
• Read a fascinating article on Tablet Computing - entitled Tomorrow's Laptops - Pen, Voice, Touch Input, New Displays, New Chips, New Storage Devices, New Visions: ... using the system and its controls is difficult to describe, perhaps because it is so natural ... this is where Slate is headed ... Ease of Use is Key ... - on and on in a 'State of the Art' section. All from BYTE Feburary 1991 - go Go ! (as it were).
• Poked the reprap idly, hmm - where did all the time go that I used to have, to play with it: LibreOffice ? Sermon from Park Street in the evening on Judges - and the calling of Gideon.

• Lie in, set too with the superdeck paint on the (already mouldering) new front side gate. Looks a lot better now. Lunch. Onto paint chipping on the stairs, putting another coat onto the stair-well (finally looking quite pleasant), and then painting the back of the workshop.
• Sat around reading in the afternoon, while the babed played, trampolined a little; read stories, popped everyone to bed. Hacked on this and that in bed - got the broffice splash re-work finally done and working.

• To work, quick mail skim, dipped into fixing a silly in the code, polished LXF column. Plugged through Norbert's nice FASTBOOL patches doing spot sanity checks; looking good, fixed a real nasty in the editeng using a 'FASTBOOL' as an integer loop iterator.
• Added a SAL_N_ELEMENTS macro to sal (long overdue), and added a new easy hack to use it widely. Also added a minutes. Dug at some code with Caolan, merged patches, fun.

• Up late, prodded mail, really pleased to see that Gianugo has been hired by Microsoft - they couldn't hope for better advice on Free Software.
• Write notes, pondered Linux Format column, prodded bugs, chat with Chen. Added a tweak to remind people to source ooenv having done a linkoo - to avoid unexpected startup failures.

• Up early, off to school with the babes, and to Burger King for a pick-up from Nicola to TEAM. Lots of interesting study, started the Bible Overview proper in Genesis, studied tools and techniques for understanding Epistles, with reference to Galations (with Eden Baptists' pastor).
• Home, conference call with the lads; off into Cambridge for Rob's twenty-seventh' birthday party; fun - good to meet his crowd of friends, bus back rather late - a fatality on the train line.
• Pleased to see the LibreOffice Planet spring into being, and looking pretty.

• Prodded mail, and more patches. Call with fcrozat. Plugged away at some admin pieces. Sad to see Ari leave Nokia.
• Chased an e-d-s bug, filed a trace and comments; JP's staff meeting. Call with Pete & a partner, then with Jared.
• Created a couple of LibreOffice accounts, and dug through the admin backlog looking for low hanging fruit to fix. Bed late.

• Up early, dropped babes at school, wandered through E-mail. Admired the large patches Norbert Thiebaud is creating, and set too writing a vtable enumeration tool for OpenOffice. But why ?

Re-factoring C++ and the vtable problem

The opportunity

Consider that you have large piece of C++, over many years, for reasons unknown, it has accumulated (at least) four boolean types: FASTBOOL, BOOL, sal_Bool and of course the native boolean type bool. Now consider a class with a virtual method in it:

class Base {       ... virtual void setValue (BOOL b); /* A */ }
class Sub : Base { ... virtual void setValue (BOOL b); /* B */ }

Please notice that these two classes are highly unlikely to be in the same header file, and may even be separated by many leagues of code, so you might never notice.

A small cleanup

Now imagine that we want to cleanup this obsolete 'BOOL' to a 'bool' - that would be nice and pretty; so - since we are hacking on 'Sub' - we change it there, and we change all the callers in Sub's implementation:

class Base {       ... virtual void setValue (BOOL b); /* A */ }
class Sub : Base { ... virtual void setValue (bool b); /* B2 */ }

At this point we have a problem - we suddenly broke the functionality of the Base class, and anyone calling the virtual method A. This is because without realising it we introduced an entirely new method with the same name and a different signature - by the miracle of polymorphism. ie. now we have two virtual 'setValue' methods, and B2 does not override A. That is not good.

Of course, because this is a feature, the compile will just love it, and suddenly, magically compile a completely different piece of code, and better - since the change is almost invisible to the casual reader this will be a nightmare to debug. Interestingly changes such as the above may seem rare, but are only the top of the iceberg - a more common incidence of these crops up in 64bit porting, where there is some mix of different types down the virtual function chain - someone overrides with a different typedef - all of which previously resolved to 'int'.

How to avoid some grief

The (rather dumb) tool I hacked up today tries to avoid this sort of problem. It does this by noticing that when we make this sort of slip, we add a new virtual method to Sub's v-table. Thus - if we simply print-out all the vtables sizes (in slots) before, and then after a change, we can rapidly check if they match. If there is a disparity - we can find the class that changed size, wind back to (all of) its parents, and work out which one is missing a suitable change. This technique was remarkably succcessful in finding 64bit portability problems (by comparing 64bit vs. 32bit builds) when Kendy was working on the original OpenOffice.org port to 64bit architectures. The tool is here

How to re-introduce grief

One final deep joy about C++ is that there is no real need to mark an overriding method as virtual (though most people, most of the time tend to). So we can write the original problem just as well as:

class Base { ... virtual void setValue (BOOL b); /* A */ }
class Sub : Base {    ... void setValue (BOOL b); /* B3 */ }

It is worth noting at this point, C#'s much better, and more readable syntax, mandating an 'override' for B3. Then lets re-add the great change we made before:

class Base { ... virtual void setValue (BOOL b); /* A */ }
class Sub : Base {    ... void setValue (bool b); /* B4 */ }

So - now we have an even nastier beastie - in this case by changing the type, we suddenly stop overriding A, but instead of creating a new virtual method, we create a normal method B4. Worse - this doesn't change the size of the vtable - so it is even harder to notice. At this point, at the moment - you loose.

All is not lost (in future)

Fortunately, this shouldn't be entirely the end of the story; in subsequent iterations, it should be easy to add a 'method count' that can be calculated for each class - and compared before and after, this would allow B4 to be detected rather easily. Perhaps tomorrow I'll expand the tool to do that too. Probably I am missing some other obvious gotchas, but getting this wrong - particularly in deeply inherited hierarchies, with complex interactions can create some particularly nasty problems to debug.

• Up late, sync, with JP, prodded mail, patches and blog entries.
• Update1: Mikael Olenfalk points out that the Mozilla guys have some magic keyword they use for this, along with Dehydra - probably libreoffice should be doing the same; we could add an override keyword that could be enforced by Dehydra, and #define it to virtual + some attribute (I guess).

• Update2: A friendly hacker advised me that the Sun Studio compiler warns about virtual functions being hidden by non-virtual ones, it seems g++ has an equivalent -Woverloaded-virtual -Werror=overloaded-virtual that should be used to detect this case. There is also a nice wikipedia page with the latest C++ standards thinking

• Up earlyish, screwed the finished articles to the ceiling. Father left for Wales, read stories to H. until lunch.
• Light lunch, took three smaller babes off to Sarah's baptism at Isleham a jolly time. Back, tea for babes, packed them to bed, and slept ourselves.

• Up earlyish, dealt with babes while J. slept. Poked idly at the reprap - still not printing gears anything like as good as it used to (possibly the fault of the gears ?).
• Off to Ridgeons with Father to buy fixings and paint; I hate haggling for goods - the Quakers introduced honest pricing in the 17th century - but it still hasn't reached UK builders merchants; despite getting a ~50% discount, I still got over-charged, annoying.
• Back home, set too cleaning the chimney - lots and lots of soot, straining with the brushes; he who has clean hands and a pure heart ? - after removing the soot to the bin, extensive showering and cleaning of the front room: fires possible once more.
• Late lunch, J. returned with H. and M. from a party. Set too with putting up some second-hand blinds, unfortunately too large, cutting down possible, but required metal-work, re-stringing etc. Became somewhat expert (over the next hours), in the (re-)construction of cheap wooden venetian blinds. H. sick; cleaned that up. Bed late.

• A new month; walked the babes to school in new shoes; some not inconsiderable pain in the heels - a little reminiscent of suddenly having to make libreoffice /actually/ build for lots of people, on different systems rather suddenly - and without any excuses.
• Plugged away at the binfilter - unfortunately, corrupting memory in destructors called at exit. Dis-satisfied with intermittent brokenness around heap corruption, fixed (and deferred) the svx part also trivially accelerating startup, Caolan has some similar mutex / thread related crash fixes to go in too. Re-started the build.
• Lunch, came back to a build - the joys of a 6 CPU icecream cluster. Fixed python sillies for installation using system-python (so gdb works), and got a nice working, debuggable build; great. Updated the website to build like that, and without the (now fixed) binfilter. Tested equivalent svx crash-on-exit fix, and pushed it.
• Tweaked the LibreOffice easy hacks page, and added a few more useful pieces. Plugged through a 'self review' in some infernal internal personel management web tool.