Friday, 2007-05-04

*** Pseudonym has joined #Aqsis00:09
*** render has quit IRC01:03
*** scorp007 has joined #aqsis04:03
scorp007hi04:04
*** ShortWave has quit IRC04:29
*** ShortWave has joined #aqsis04:30
scorp007hi ShortWave04:36
ShortWavehiya04:36
scorp007how's development on Neqsus going?04:37
ShortWaveit's going04:37
scorp007well, I assume?04:37
ShortWavefor the moment04:37
ShortWaveI think I'm going to hit a brick wall soon, where I have to wait on the next blender release tho04:38
scorp007ok -- do you basically generate a rib on the fly, then invoke the render command as a separate process? Or do you link directly against the library?04:38
ShortWaveoption #104:39
scorp007well, you don't have to wait for the next release, you can stay up to date with the cvs04:39
ShortWavewell04:39
ShortWavethe CVS is less than stable04:39
scorp007hmmm... anyone good with latex here?04:44
scorp007nevermind, worked it out05:41
*** ShortWave has quit IRC06:16
*** Pseudonym has quit IRC06:23
*** Pseudonym has joined #Aqsis06:28
*** render has joined #aqsis07:26
*** pgregory has joined #aqsis07:48
*** ChanServ sets mode: +o pgregory07:48
pgregorymorning all07:57
*** pgregory sets mode: +vvv scorp007 render Pseudonym07:57
PseudonymG'day.07:57
pgregoryhi Andrew07:58
pgregorywas thinking about that binary/XML thing a little07:58
*** joeedh is now known as joeedh_walk07:58
PseudonymYeah?08:00
pgregoryhow does it work in the case of different content types?08:01
pgregoryi.e. we might have a buffer full of chars, shorts, longs or floats.08:01
PseudonymGood question.08:01
pgregoryif they are chars, for examples, surely we don't actually want to do any reordering at all.08:01
pgregory?08:02
PseudonymSure.08:02
PseudonymWell, htons() and ntohs() work on "shorts".08:02
Pseudonymand htonl() etc work on "longs".08:02
Pseudonyms == short, l == long08:02
pgregoryyeah08:02
pgregorypresumably, a float can be treated as a long?08:02
PseudonymPresumably, yes.08:02
PseudonymVia a packed C union, for example.08:02
Pseudonymdouble would be a bit tricker.08:03
pgregorywe don't deal with doubles in the display interface08:03
PseudonymRight.08:03
pgregoryok, looks like I need to take the elementsize into consideration when packing the data.08:05
PseudonymRight.08:05
pgregorynot too bad, I just hope all this packing/encoding doesn't hit the transfer time too much.08:06
pgregorymight need to look into threading to handle this.08:06
pgregoryalso, if you've got a moment, would appreciate your thoughts on another part of this, code design related.08:07
PseudonymOK08:09
PseudonymShoot.08:09
pgregoryright, the XML packets for the messages are modelled very heavily on the Dspy interface data, making them a simple 'forwarding' mechanism.08:09
PseudonymSure.08:10
pgregoryso at the moment, I simply package up the data passed into the DspyImageOpen function into an XML packet.08:10
pgregorywithin the DspyImageOpen function.08:10
PseudonymRight.08:10
pgregorybut I'd like to separate this functionality a bit, message classes semm overkill, as they would be transient, resulting in effectively a useless class with a constructor that builds the message and sends it.08:11
pgregorybut I'd still like to have it abstracted a bit for maintenance reasons, I have a couple of ideas, wondered if you had any thoughts.08:11
PseudonymLet me check one thing.08:12
PseudonymI assume the new fancy framebuffer is in two parts.08:12
PseudonymOne a Dspy shared object, the other the GUI code, right?08:12
pgregoryright08:12
PseudonymOK, just checking.08:12
pgregorythe Dspy server is very lightweight, or will be, just packaging up the data into XML and sending it off to the app.08:13
PseudonymSure.08:13
PseudonymHmmm.08:13
PseudonymHow are you packaging XML?08:13
pgregoryTinyXML08:13
PseudonymAre you using some standard XML library?08:13
PseudonymRight.08:13
PseudonymSo you're building an XML tree, sending it, then destroying the tree.08:13
pgregoryyes08:13
PseudonymSeems to me that it's only once per frame.08:13
PseudonymIs it worth optimising?08:14
pgregoryfor the open, yes, but not for the data.08:14
pgregoryI'm not really worried at this point about optimisation, just clean coding.08:14
PseudonymRight.08:14
pgregorydo you think it's reasonable to just have the XML build/send/teardown in the Dspy??? functions?08:15
PseudonymI don't see why not.08:15
pgregorywith no further abstraction? if so, I'm happy to go with that, just wanted to find the 'right' solution.08:15
PseudonymYeah, it seems fine, especially for now.08:15
pgregorycool, glad you said that, that's what I was leaning to.08:16
Pseudonym:-)08:16
PseudonymIt's simple, and it works.08:16
pgregoryall other abstraction methods I've thought of seemed over complicated for the task, hiding for the sake of hiding.08:16
PseudonymYeah.08:16
*** joeedh_walk is now known as joeedh08:16
PseudonymIt's just a stub.08:16
PseudonymI don't think there's any point building internal abstraction layers.08:17
PseudonymEspecially if TinyXML is easy to use.08:17
pgregoryvery easy/clean08:17
*** Anteru has joined #aqsis08:26
AnteruMorning08:26
pgregorymorning Anteru08:27
PseudonymG'day.08:27
*** pgregory sets mode: +v Anteru08:27
AnteruWhat's up?08:30
AnteruHow should I progress with program_options ... ? I can work on it locally with Boost 1.34 beta but it will break any pre-1.34 build with VC8 :/08:30
pgregorysounds like a good reason to stick with what we've got, especially as it works.08:31
pgregoryit isn't broken, it doesn't have any known problems/weaknesses, so why replace it?08:31
scorp007sounds like a bug in boost. (Or perhaps they made a workaround for VC8 in the beta.)08:32
AnteruNo, it's a bug in VC8 being too aggressive in their checks ...08:32
AnteruMore or less08:32
scorp007I personally don't see how being 'too strict' is a bug -- unless they violate the standard in doing so.08:34
pgregoryAnteru: what problem does using boost::program_options solve?08:34
*** AlexK has joined #aqsis08:34
PseudonymThere's a new Boost release soon, isn't there?08:35
scorp007pgregory, my guess would be potentially cleaner code, in not having to reinvent the wheel. I could be wrong though...08:35
scorp007pgregory, although, if it involves existing working code, that may not outweigh it's advantage.08:36
scorp007involves rewriting*08:36
pgregorywe're not 'reinventing' the wheel, we already have a wheel.08:37
PseudonymSeems to me that rewriting existing code plus breaking compilers that we use is a lose-lose.08:37
scorp007right.08:37
pgregoryagreed08:37
scorp007if we didn't have a wheel, it might have made more sense08:37
PseudonymIf it wasn't for the broken compiler issue... then if someone wanted to do the rewrite and make the code cleaner, well, that's cool.08:38
AnteruYeah, that was the original intention :/08:40
AnteruDidn't suspect boost 1.33.1 would break VC8 :(08:40
PseudonymGotta go.08:48
PseudonymNight everyone.08:48
AnteruBye08:48
*** Pseudonym has quit IRC08:48
*** cgTobi has joined #Aqsis08:50
*** ChanServ sets mode: +o cgTobi08:50
cgTobimorning all09:10
*** Anteru has quit IRC09:21
*** renderguy has joined #aqsis09:53
renderguyMorning all09:53
renderguy.TGIF!09:53
*** AlexK has quit IRC09:59
cgTobiHi LEon10:28
*** cgTobi sets mode: +v renderguy10:28
renderguycgTobi: Moin, how is/was German?10:41
renderguycgTobi: German == Germany.10:41
cgTobi127.0.0.1 sweet 127.0.0.110:42
cgTobi:)10:43
renderguycgTobi: :-)11:00
*** Anteru has joined #aqsis11:01
*** Joron has joined #Aqsis11:30
renderguyJoron: Morning.11:33
JoronHello11:34
cgTobiLeon, have you tried to build svn 1020?11:34
renderguycgTobi: Not yet, why?11:51
cgTobiI have problems with mkstemp()11:53
cgTobijust wanted you to confirm it11:53
cgTobias soon as I switch back to mktemp() it works fine.11:53
cgTobiI need to investigate that after lunch11:53
cgTobibbl11:53
renderguycgTobi: Which OS is that issue under?11:54
*** Anteru has quit IRC12:06
*** Joron has quit IRC12:14
cgTobirenderguy: OSX12:22
renderguycgTobi: What error do you get (message)?12:27
* renderguy notices the Win32 build seems to be okay, though Linux also has issues now.12:28
cgTobibuild/shadercompiler/aqsl/aqslcomp.cpp:163: error: invalid conversion from 'int' to 'char*'12:29
renderguycgTobi: Yep, same issue on Linux.12:29
cgTobiAh, IC12:30
renderguyAnyone know the difference between 'mktemp()' and 'mkstemp()'?12:30
pgregorysecurity on Windows I think, I thought AndyB's change should only have affected Windows.12:31
renderguypgregory: Quite the opposite.  ;-)12:31
cgTobinot really :)12:33
cgTobi#ifdef  AQSIS_SYSTEM_WIN3212:33
cgTobi                                tempname = _mktemp( ifile );12:33
cgTobi#else12:33
cgTobi                                tempname = mkstemp( ifile );12:33
cgTobi#endif //AQSIS_SYSTEM_WIN3212:33
pgregoryoops, my mistake12:33
pgregoryaha, this is why I hadn't used mkstemp before (I do wish people would read the logs before making changes).12:34
renderguycgTobi: Is that your fix or what's currently in the code?12:35
pgregorymktemp returns the name of a temporary file, mkstemp creates the file, which is no good to us, as we need the name to pass through to the preprocessore.12:35
cgTobirenderguy: that's the current code12:35
pgregorythat change needs to be rolled back12:35
renderguypgregory: Perhaps using 'tempfile()' would be a better soltion?12:37
renderguypgregory: http://unixhelp.ed.ac.uk/CGI/man-cgi?tempfile+112:38
pgregorya) tempfile is a tool, not a function, b) it still creates a file, we need a filename.12:39
renderguypgregory: My mistake, what is wrong with 'mktemp()' anyway?12:41
pgregorythere is a known race condition.12:41
pgregoryif you call mktemp, get a 'safe' filename back, and before you create the file, another process calls mktemp and gets the same 'safe' filename back, you're stuffed.12:42
renderguypgregory: K, but has it ever caused us an issue yet?12:43
pgregoryno12:43
renderguypgregory: Guess we have our fix/answer then.  ;-)12:44
pgregorythe 'fix' is to change the way we do preprocessing, but as our preprocessing is based on a thirdparty preprocessor, it's difficult to make the changes we need.12:45
*** Auralis has quit IRC12:49
*** Auralis has joined #aqsis12:49
pgregoryit is a definite problem, and something we should be wary of, and change if/when we can.12:50
*** cgTobi sets mode: +v Auralis12:52
*** mafm has joined #aqsis13:16
mafmhi13:16
pgregoryhi mafm13:20
*** pgregory sets mode: +v mafm13:20
renderguymafm: Afternoon.13:23
mafm:)13:30
renderguymafm: How are you projects going?13:37
mafmwhich projects, do you mean school projects?13:45
renderguymafm: Yep.13:46
mafmI'm trying to finish a compiler for one of the subjects and that's almost finished13:46
mafmbut the other ones are delayed :S13:46
mafmI think that I'll have to leave some of the projects/subjects for a better occasion13:48
mafmbtw, I don't know whether there are things which need preparation before SOC projects start, besides the SVN account?13:54
pgregorywe'll need to setup a branch for you to work in, although you should be able to do that yourself if you want, once you're commit authorised.14:00
mafmI mean if there's anything else to do before I start? I don't think so14:02
*** Anteru has joined #aqsis14:03
pgregoryno, don't think so14:04
mafmgoody14:08
*** cgTobi has quit IRC14:22
pgregoryWARNING: new-framebuffer branch is broken, don't update!14:50
pgregorygotta go, cya all later.14:50
*** pgregory has left #aqsis14:50
*** scorp007 has quit IRC14:55
renderguytcolgate: You around?14:57
mafmI haven't seem him on-line for weeks15:27
renderguymafm: He must be busy, ill or on holiday.15:36
mafmhole that it's the latter :)15:37
mafmI have to go now, see you15:37
*** mafm has quit IRC15:37
*** Anteru has quit IRC15:43
*** renderguy has quit IRC16:11
*** Anteru has joined #aqsis17:26
*** cgTobi has joined #Aqsis17:38
*** ChanServ sets mode: +o cgTobi17:38
*** Anteru has quit IRC17:54
*** cgTobi has quit IRC17:55
*** joeedh is now known as joeedh_biking20:48
*** joeedh_biking is now known as joeedh21:08
*** cgTobi has joined #Aqsis22:47
*** ChanServ sets mode: +o cgTobi22:47
*** cgTobi has quit IRC22:47
*** render has quit IRC23:16
*** joeedh is now known as joeedh_class23:16
*** render has joined #aqsis23:17

Generated by irclog2html.py 2.5 by Marius Gedminas - find it at mg.pov.lt!