Older blog entries for redi (starting at number 111)

chalst said "more freedom and resources when creating your department"

I understand this was definitely one of the reasons in his case. I confess my source for this was an overheard dinner conversation between Mrs S and, IIRC, Peter Sommerlad (another academic working with C++.)

cdfrey, it might be that some graph has to be recalculated (maybe overnight) before it actually takes effect.

There are definitely people who I don't see on recentlog because I've rated them 2, and I haven't cert'd them directly. But maybe you're right, and they just happen to be below 3 for me because someone I've cert'd has rated them low.

Update: no, it's instant ... I've just rated arthur and the entry below this one has disappeared. You are browsing at thresh!=0, right?

Upupdate: no, it's not instant! I experimented with rating mattl at 2, and I still see my previous rating next to his username, and still see him on the recentlog at thresh=3. Even if I certify mattl, I still see my old diary rating. Curioser and curioser.

cdfrey, you don't have to certify someone to rate their diary, which I think lessens some of the other problems you describe. If I rate someone 2 and they disappear for you at your preferred threshold, it's easy for you to rate them higher again. You don't need to certify them to do that, but you do need to read recentlog without a threshold now and then, to see if there's anything you're missing.

ncm, I was trying to explain the same ideas at work yesterday (particularly the destructors aspect) but didn't say it nearly as well. I also quoted "[inlines] are the third most-misused C++ feature (after inheritance and overloading)" from your C++ Style Guidelines, so I shall just resort to parrotting you in future :-)

A recurring theme in the C++ talks I gave at my previous job was that basic language features such as destructors are far more important than most people realise, but I'm not sure the notes capture that; I should try to expand the notes to cover more of what was discussed in the talks.

Boost and std::lib implementations apart, what examples of high-quality open-source C++ are there? Unfortunately I've not seen a lot that I'd consider worth studying, which is a real shame. Too many coders still treat C++ as nothing more than a better C, which means they write C with a sprinkling of inheritance and virtual functions, and no fewer memory leaks or broken invariants.

Quick, one more spam rating and informationseo gets deleted - kill the spammer! KILL!

Update: one less spammer.

avriettea, man section 8 is used for system admin commands by BSD-derived systems (including Darwin) and Linux, so sudo's man page is supposed to be in section 8. Some systems use section 1M for that, but you can't blame Darwin for following a longstanding BSD tradition. You may continue ranting now.

I'm glad to see that Experimental C++0x support in GCC is coming along pretty well, but also glad it hasn't all been done before I updated my GCC copyright assignment, so I won't miss all the fun :-)

I've nearly finished implementing the changes from Improving shared_ptr for C++0x for libstdc++. Variadic templates and rvalue references are the mutt's nuts, C++ library writers have never had it so good!

Also working on a non-threadsafe shared_ptr clone that can be used for objects that are ref-counted but not shared between threads (one of std::shared_ptr's weaknesses,) using the parameterised locking policy that Phillip Jordan added for GSoC 2006. With some simple changes (and additional template specialisations) I see a 20-25% speed-up for shared_ptr operations in multi-threaded code (there should be no improvement for single-threaded code, since atomics are only used when multiple threads exist.) I think this would make a nice extension for libstdc++, but if not I will try to get the specialisations in so that the faster shared_ptr can be implemented entirely in user code.

While I agree with fzort about recentlog, I think it's better than if advogato had disappeared entirely. The diary ratings still work well for me; even when I haven't rated someone's diary myself, the people I trust (i.e. those I've cert'd) seem to be interested in topics that interest me, so their ratings work for me too.

apenwarr is usually interesting. (Update: that should really say "is almost always very interesting.") The next version of the C++ standard introduces the idea of a constexpr function, which is a function that can be used where an integral-constant-expression is needed, which will allow some uses of the preprocessor to be eliminated (e.g. using the maximum of two constants as an array bound) ... but doesn't help with the good examples Avery gives.

robogato, the new account page tells spammers that "Links on new user profile pages are suppressed altogether until your account is certified by a trusted user" which doesn't seem to be the case for the Homepage URL, should it be?

Thinking about spammers who create advo accounts to push their business, such as tabernae: Maybe after an account gets deleted as spam, any links they added to their account could be temporarily added to a wall of shame page here on advogato, which would link to their site with some suggestion of narcotics out-sourcing or kiddie porn solutions. Those who look for spammers in the "recent people joining" list should remember not to click the links on spammers' profiles, so they don't get Referer logs that show spamming advogato brought them traffic; if you want to check the authenticity of the user, copy'n'paste their link instead.

P.S. chinatoper was another damn spammer, who seems to trade in low-quality tat and used toilet paper (don't click those last two links)

[censored grumpy meme rant!]

But then I'm grumpy today. I don't have anything better to say either.

PStreams is being used by s11n to serialise C++ objects over email! (and other, more widely useful media too.) A friend is also about to publish a genetic algorithms project that uses pstreams+ssh for communication between nodes of a cluster.

Earlier this year I announced I'd be winding up the mysqlcppapi project I took over from murrayc, since the only activity was me rejecting all the spam that gets sent to the mailing list. Some users came out of the woodwork to ask that the project stay alive and as the quantity of spam fell for a while, I gave it a reprieve. The quantity of spam has climbed back up and I'm thinking of dropping the project again (of course anyone's welcome to continue development if they want, it's open source after all.)

The FSF should have received my updated copyright assignment for GCC by now, although given the UK's recent postal strikes it might still be in London somewhere.

The project URLs look fine now. Thanks, Steven!

There seems to be a bug in mod_virgule's handling of URLs on project pages, see the homepage link for WvStreams. The ? character is turned into a ridiculously long representation of 0.0