| --- Log opened Sat May 07 00:00:45 2016 |
| 00:29 | <&McMartin> | Man |
| 00:29 | <&McMartin> | I am tempted to set up my system to play the MoO2 new-technology jingle whenever my unit tests pass |
| 00:32 | <~Vornicus> | At resmark we used the Sims "water accident" song whenever somebody pushed a commit that failed |
| 00:41 | < catalyst> | I got the unit tests building and passing again on $project today |
| 00:41 | < catalyst> | they tend to rot when I'm not there :< |
| 00:44 | <&ToxicFrog> | Boo |
| 00:44 | <&ToxicFrog> | You need to enforce test greenness with nuclear failsafes^W^W presubmit checks. |
| 00:46 | <&McMartin> | It is not always obvious in our cases what correct behavior even is yet, so things are still in flux |
| 00:47 | <&ToxicFrog> | McMartin: the rule in these parts is that you don't have to have unit tests (although if you don't, be prepared to convince the code reviewer why this is OK) |
| 00:47 | <&McMartin> | Also, a lot of my "unit tests" right now are actually things like "ensure the language runtime actually destroys objects in the order the documentation implies but does not actually *state* that it does" |
| 00:48 | <&ToxicFrog> | But you can't check anything that breaks those tests that already exist. |
| 00:48 | <&McMartin> | Yeah |
| 00:48 | <&McMartin> | This implies things like "a meaningful unit test infrastructure that can be automatically run", or, if we're going to get right down to it, "a centralized build system" |
| 00:48 | <&ToxicFrog> | (or well, you can, but it requires pushing the Big Red Button and you'd better have a really good reason for it) |
| 00:48 | <&ToxicFrog> | It does, yews. |
| 00:48 | <&McMartin> | The project I'm on is... not at that level of development yet. |
| 00:49 | <&McMartin> | We've finally managed to set up a system that will properly do unit tests in an iOS simulator but (a) that's not *really* testing the code and (b) it still only really works on his rig. |
| 00:51 | <&McMartin> | But yeah, my region of responsibility right now is a library with an API that is still being negotiated |
| 00:52 | <&McMartin> | So my "unit tests" at the moment are mostly 20-line sample programs that print out fixed values |
| 00:52 | | * ToxicFrog nods |
| 00:52 | <&ToxicFrog> | catalyst, on the other hand, works for a long-established game developer~ |
| 00:52 | <&McMartin> | One could imagine feeding that to a diff on every compile, but, well, with no buildslaves this is literally more work than just typing a command after a build. |
| 00:53 | <&McMartin> | Yeeees |
| 00:53 | <&McMartin> | Also, eventually the library will have actual enforceable contracts, and a bunch of good stuff is waiting to go on that~ |
| 00:53 | <&McMartin> | I still want to see them running Actually On An iOS Device though before I'm comfortable with it |
| 00:53 | <&McMartin> | ... especially since I'm going to be relying on std::is_standard_layout<> here |
| 00:57 | <&McMartin> | Which in turn may rely on several reinterpret_cast<>s, unless I can actually use static_cast<> |
| 00:57 | | * McMartin checks those rules. |
| 00:57 | <&McMartin> | Anyway. Stuff where architectural trivia is likely to introduce Surprises(tm). |
| 01:01 | <&McMartin> | Yeah. I can't use static_cast. That is... sensible but makes me feel like there has to be a better way to do this than a pair of reinterpret_cast<>s. |
| 01:02 | <&McMartin> | I have an object of unknown type, with a member of known type and a known offset within said unknown type. I want a pointer to that member. |
| 01:03 | <&McMartin> | The best technique I have for that right now is to reinterpret_cast<char *> it, add the offset, then reinterpret_cast<known_type *> the result. |
| 01:04 | <&McMartin> | Maybe I just want to use pointers to members directly. |
| 01:05 | <&McMartin> | ... but maybe I *don't* because then I need a separate object for representing each different known type. |
| 01:05 | <&McMartin> | (I'm trying to do a semi-general demarshaller for a passel of related types) |
| 01:09 | <&McMartin> | Yeah, I think if I want a consistent object descriptor I need to rely on offsetof (which is OK because std::is_standard_layout<> says it is) and then reinterpret-through-char * to get at the member. |
| 01:12 | <~Vornicus> | I failed miserably at understanding the term "demarshaller" and thought you were speaking like yiddish for a minute |
| 02:13 | | Crossfire [Z@Nightstar-r9lk5l.cust.comxnet.dk] has quit [Ping timeout: 121 seconds] |
| 02:26 | | catalyst [catalyst@Nightstar-890.fhp.187.81.IP] has quit [[NS] Quit: Leaving] |
| 02:28 | | catadroid [catalyst@Nightstar-63m6gl.dab.02.net] has joined #code |
| 03:35 | <~Vornicus> | dem-ars-hal-ler or something |
| 03:47 | <&Derakon> | Dem arshes. |
| 03:47 | <&Derakon> | Sean Connery admiring the scenery. |
| 05:23 | | Derakon is now known as Derakon[AFK] |
| 06:12 | | celticminstrel [celticminst@Nightstar-q0f7bb.dsl.bell.ca] has quit [[NS] Quit: And lo! The computer falls into a deep sleep, to awake again some other day!] |
| 06:43 | | Alek [Alek@Nightstar-9qtiqv.il.comcast.net] has quit [[NS] Quit: brb] |
| 06:53 | | Alek [Alek@Nightstar-9qtiqv.il.comcast.net] has joined #code |
| 06:53 | | mode/#code [+o Alek] by ChanServ |
| 06:55 | | Alek [Alek@Nightstar-9qtiqv.il.comcast.net] has quit [[NS] Quit: ] |
| 06:57 | | Alek [Alek@Nightstar-9qtiqv.il.comcast.net] has joined #code |
| 06:57 | | mode/#code [+o Alek] by ChanServ |
| 07:06 | | himi [fow035@Nightstar-v37cpe.internode.on.net] has joined #code |
| 07:06 | | mode/#code [+o himi] by ChanServ |
| 07:56 | | Kindamoody[zZz] is now known as Kindamoody |
| 07:57 | | catadroid` [catalyst@Nightstar-rurha2.dab.02.net] has joined #code |
| 08:00 | | catadroid [catalyst@Nightstar-63m6gl.dab.02.net] has quit [Ping timeout: 121 seconds] |
| 10:00 | | Crossfire [Z@Nightstar-r9lk5l.cust.comxnet.dk] has joined #code |
| 10:00 | | mode/#code [+o Crossfire] by ChanServ |
| 10:05 | | catadroid` [catalyst@Nightstar-rurha2.dab.02.net] has quit [[NS] Quit: Bye] |
| 10:43 | | Vornicus [Vorn@ServerAdministrator.Nightstar.Net] has quit [Connection reset by peer] |
| 10:44 | | catalyst [catalyst@Nightstar-bt5k4h.81.in-addr.arpa] has joined #code |
| 12:08 | | Kindamoody is now known as Kindamoody|afk |
| 12:13 | | celticminstrel [celticminst@Nightstar-q0f7bb.dsl.bell.ca] has joined #code |
| 12:13 | | mode/#code [+o celticminstrel] by ChanServ |
| 12:28 | | Reiver [quassel@Nightstar-ksqup0.co.uk] has quit [Connection closed] |
| 12:28 | | Orthia [quassel@Nightstar-ksqup0.co.uk] has quit [Connection closed] |
| 12:28 | | Orthia [quassel@Nightstar-ksqup0.co.uk] has joined #code |
| 12:28 | | mode/#code [+o Orthia] by ChanServ |
| 12:28 | | Reiver [quassel@Nightstar-ksqup0.co.uk] has joined #code |
| 12:28 | | mode/#code [+ao Reiver Reiver] by ChanServ |
| 15:25 | | catalyst [catalyst@Nightstar-bt5k4h.81.in-addr.arpa] has quit [[NS] Quit: Leaving] |
| 15:28 | | thalass [thalass@Nightstar-283.o7s.158.104.IP] has joined #code |
| 15:28 | | mode/#code [+o thalass] by ChanServ |
| 15:30 | | catalyst [catalyst@Nightstar-bt5k4h.81.in-addr.arpa] has joined #code |
| 16:20 | | Crossfire [Z@Nightstar-r9lk5l.cust.comxnet.dk] has quit [Ping timeout: 121 seconds] |
| 17:06 | | Shady [ShadyGuru@Nightstar-ut3scg.t132.ptd.net] has joined #code |
| 17:25 | | Shady [ShadyGuru@Nightstar-ut3scg.t132.ptd.net] has quit [[NS] Quit: Leaving] |
| 17:54 | < simon_> | what's a lossless decimal type? googling gives me Data.Decimal, but I don't know if it's standard. |
| 17:57 | < catalyst> | fixed point is a good term to look for |
| 18:00 | <&ToxicFrog> | simon_: what do you mean by "lossless" here? |
| 18:00 | <&ToxicFrog> | Fixed point is lossless inside its range, but has limited range (for that matter, floating point is also lossless under the right conditions) |
| 18:03 | <&ToxicFrog> | There's "can express any rational number", which is a different problem and not the same as "can express numbers to arbitrary precision", for which you want an arbitrary-precision/infinite-precision/bignum type. |
| 18:09 | < simon_> | ToxicFrog, whoops, thought this was #haskell :) |
| 18:10 | < simon_> | ToxicFrog, it can be bounded, but it shouldn't mess with the fractions like floats. |
| 18:22 | <&ToxicFrog> | Floats don't mess with the fractions either as long as you stay within the bounds! |
| 18:22 | <&ToxicFrog> | So the question here really is "what kind of behaviour do you want when you exceed the precision bounds of the type" |
| 18:55 | | thalass [thalass@Nightstar-283.o7s.158.104.IP] has quit [Ping timeout: 121 seconds] |
| 19:00 | | Derakon[AFK] is now known as Derakon |
| 19:17 | | thalass [thalass@Nightstar-283.o7s.158.104.IP] has joined #code |
| 19:17 | | mode/#code [+o thalass] by ChanServ |
| 19:19 | | thalass is now known as Thalass|blendy |
| 19:20 | <@celticminstrel> | I wouldn't say bounds is the criteria for floats not messing with fractions. |
| 19:20 | <@celticminstrel> | It depends entirely on the denominator. |
| 19:21 | <@celticminstrel> | Data.Decimal is probably an arbitrary-precision type. |
| 19:22 | <@celticminstrel> | I think even arbitrary-precision can't be lossless though - it still won't be able to represent 1/3 exactly, right? |
| 20:05 | | Vornicus [Vorn@ServerAdministrator.Nightstar.Net] has joined #code |
| 20:05 | | mode/#code [+qo Vornicus Vornicus] by ChanServ |
| 20:05 | | Vornotron [Vorn@ServerAdministrator.Nightstar.Net] has joined #code |
| 20:10 | | Turaiel[Offline] [Brandon@Nightstar-7mqsi0.mi.comcast.net] has quit [Ping timeout: 121 seconds] |
| 20:11 | | Turaiel[Offline] [Brandon@Nightstar-7mqsi0.mi.comcast.net] has joined #code |
| 20:12 | | Vornotron [Vorn@ServerAdministrator.Nightstar.Net] has quit [[NS] Quit: Leaving] |
| 20:16 | <&ToxicFrog> | celticminstrel: Data.Decimal is, according to the documentation, not arbitrary-precision. |
| 20:17 | <&ToxicFrog> | And the whole point of floating point is that whether you lose precision depends not on the denominator but on how many significant bits you need. |
| 20:18 | <&ToxicFrog> | A double can exactly represent both 2^100 and 1/2^100, but not either of those + 1 |
| 20:19 | <@celticminstrel> | But it also can't exactly represent any fraction whose denominator cannot be factored into powers of 2. (Or powers of 10, if a decimal representation is being used.) |
| 20:20 | <@celticminstrel> | Or something like that. |
| 20:20 | <@celticminstrel> | That sounds wrong somehow. |
| 20:20 | <@celticminstrel> | I think it's simply that the denominator needs to be a power of the base. |
| 20:39 | <~Vornicus> | floating point numbers have a size limit on their numerator, and their denominator has to be a power of the base. |
| 20:58 | | Thalass|blendy is now known as Thalass|nomnomnom |
| 22:03 | | Kindamoody|afk is now known as Kindamoody |
| 22:18 | | thalass [thalass@Nightstar-283.o7s.158.104.IP] has joined #code |
| 22:18 | | mode/#code [+o thalass] by ChanServ |
| 22:47 | | Derakon is now known as Derakon[AFK] |
| 23:13 | | Kindamoody is now known as Kindamoody[zZz] |
| 23:24 | | Derakon[AFK] is now known as Derakon |
| 23:57 | | thalass [thalass@Nightstar-283.o7s.158.104.IP] has quit [Ping timeout: 121 seconds] |
| 23:58 | | Thalass|nomnomnom is now known as Thalass |
| --- Log closed Sun May 08 00:00:01 2016 |