RAOF | Grr. | 00:14 |
---|---|---|
RAOF | Threadsafety by making every individual component threadsafe is maybe not the best idea. | 00:40 |
racarr | RAOF: Yeah...I had this thought when I studied | 01:38 |
racarr | Chromium and they have this...uh | 01:38 |
racarr | well rather than each component is thread safe its more like components live in threads | 01:39 |
racarr | and the threads are like | 01:39 |
racarr | higher level than the components | 01:39 |
racarr | e.g. render thread, ipc thread, etc... | 01:39 |
racarr | grr sorry I can't think of how to articulate it quickly | 01:39 |
RAOF | The google juice for that would be CSP (communicating sequential processes) I believe. | 01:39 |
racarr | but it was a really different model that I thought probably made more sense | 01:39 |
racarr | yeah I mean | 01:39 |
racarr | it looked like that because the main communication between them was | 01:39 |
racarr | this fd thing but | 01:40 |
racarr | even without going full like... | 01:40 |
racarr | I mean even with still using shared memory concurrency | 01:40 |
racarr | youc an group components in a smarter way and not | 01:40 |
racarr | require thread safety on each one | 01:40 |
RAOF | Just on the interface between them, yeah. | 01:40 |
RAOF | (CSP doesn't imply not shared-memory, either) | 01:41 |
racarr | yeah thats true. | 01:42 |
racarr | Running out to Go club (Game not Lang)! | 01:42 |
RAOF | Coo! | 01:42 |
RAOF | Have fun! | 01:42 |
AlbertA | vogons: so prepping for 0.15 release, abi compliance checker complains mir client is not ABI compatible because of enum MirPixelFormat mir_pixel_formats has changed... | 02:24 |
AlbertA | any opinions on the matter? | 02:26 |
RAOF | AlbertA: False positive. | 02:27 |
RAOF | Or, rather, only code that currently fails to work will change behaviour. | 02:27 |
RAOF | (I believe what it's complaining about in the value of mir_pixel_formats has changed) | 02:27 |
AlbertA | yeah I couldn't think of anything....but it's late.... | 02:28 |
AlbertA | RAOF: right, which should be unused or just used for bounds checking | 02:28 |
RAOF | The only way code can usefully use it is still valid (as a guard value). | 02:28 |
RAOF | Correct. | 02:28 |
AlbertA | as long as it doesn't go down should be ok.... | 02:28 |
RAOF | It technically is an ABI change, as something which called mir_foo(..., mir_pixel_formats) changes behaviour from erroring to passing. | 02:29 |
RAOF | But people who write such code deserve what they get :) | 02:29 |
RAOF | Hm. | 02:29 |
AlbertA | he he yep | 02:30 |
RAOF | Actually, that could plausibly break a test-suite. | 02:30 |
RAOF | But almost certainly doesn't, and shouldn't break any code that *users* care about. | 02:31 |
RAOF | So, go ahead. | 02:31 |
AlbertA | test-suite probing behavior of mir client apis? | 02:31 |
AlbertA | with invalid inputs? | 02:31 |
AlbertA | I suppose so....but is there such a thing in the wild? | 02:32 |
RAOF | A toolkit could reasonably use that to check its handling of Mir errors, for example. | 02:32 |
AlbertA | umm true... but that's why you mock those things :) | 02:33 |
RAOF | I don't think there is such a test-suite in the wild, and I don't think that “inputs which were invalid remain invalid” is an ABI guarantee we make. | 02:33 |
duflu | AlbertA: Yep mir_pixel_formats is intended to be a value current for the code it's compiled into. One should expect (and plan for) other users to be aware of a different value for mir_pixel_formats | 02:55 |
duflu | AlbertA: Also, my MP up for review solves an issue that will block 0.15. | 02:56 |
duflu | Although I'm making great progress optimizing QtMir, that won't be finished for a while | 02:57 |
AlbertA | duflu: thanks, which MP? | 03:02 |
duflu | AlbertA: My only one up for review :) | 03:02 |
AlbertA | always triple buffers? | 03:02 |
AlbertA | duflu: ok | 03:02 |
RAOF | Man, GLibMainLoop is thorny. | 07:43 |
duflu | greyback: Fun fact: I successfully prototyped very short buffer holds in QtMir and found less than expected performance. Then looked at Mir again and found I had regressed the short hold feature back in Feb. Mir needs fixing :) | 08:59 |
greyback | duflu: em, woo! | 08:59 |
=== andyrock_ is now known as andyrock | ||
=== guest42345 is now known as guest00|AFK | ||
=== alan_g is now known as alan_g|lunch | ||
kenvandine | kgunn, with silo 0 installed, usc doesn't start, is that known? | 13:25 |
kenvandine | ERROR: /build/mir-GtufN4/mir-0.13.4+15.04.20150717/src/platform/graphics/platform_probe.cpp(59): Throw in function std::shared_ptr<mir::SharedLibrary> mir::graphics::module_for_device(const std::vector<std::shared_ptr<mir::SharedLibrary> >&) | 13:25 |
kenvandine | Dynamic exception type: N5boost16exception_detail10clone_implINS0_19error_info_injectorISt13runtime_errorEEEE | 13:25 |
kenvandine | std::exception::what: Failed to find platform for current system | 13:25 |
greyback_ | kenvandine: known issue, USC needs rebuild | 13:31 |
kenvandine | greyback_, ok, so is someone doing that? | 13:31 |
kgunn | kenvandine: yep, sorry.... | 13:31 |
greyback_ | kenvandine: yep, will ping when it ready | 13:31 |
kenvandine | thx | 13:32 |
greyback_ | someone rebuilt bits before it was all ready... :) | 13:32 |
kgunn | i rebuilt mir last night and got stuck on the u-s-c merge | 13:32 |
kgunn | greyback_: ah,....that would be me | 13:32 |
kgunn | greyback_: altho i did retarget mir0.14 and the whole bit | 13:32 |
greyback_ | kgunn: no worries, I had mir & qtmir updated, but wanted to finish usc before rebuilding | 13:33 |
kgunn | yeah, i prolly shouldn't have done the piece wise rebuild | 13:33 |
kgunn | esp after sending out mails | 13:33 |
kgunn | classic | 13:33 |
* kgunn hasn't checked mail and can only imagine | 13:34 | |
=== alan_g|lunch is now known as alan_g | ||
=== chihchun is now known as chihchun_afk | ||
=== dandrader_ is now known as dandrader | ||
=== dandrader is now known as dandrader|afk | ||
=== alan_g is now known as alan_g|EOW | ||
=== dandrader|afk is now known as dandrader |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!