text
stringlengths
1
5.2k
labels
listlengths
1
1
you need this virt driver fake in the above if block for eq
[ 0 ]
It like this Executeable (include opengl32.lib) => call the opengl32.dll(which sucks) functions => call the driver functions
[ 1 ]
and the admin could pass auth but be passing a garbage project id that is the thing we need to validate you suggest to create another user that would authenticate at keystone using username and password trip validate the project id trip i suggest to take user s token and using it validate the project id at keystone that s one less trip to keystone one less credential to manage one less thing to change in keystone
[ 0 ]
? why not something like it hums or it glows brightly or some shit
[ 1 ]
ah i was being stupid and was looking at normal route command not ip route which has the replace operation that exists even on
[ 0 ]
nit i would kill that space after the comma and just make it join tags
[ 0 ]
i think users are always more lazy than we expected evey thing can easier their life do have a value to be done
[ 0 ]
Oh sick. Nevermind then.
[ 1 ]
nit needs to be removed
[ 0 ]
super minor little nit humanizetime it may sound funny but i think of it as make time human what do you think about maketimehumanreadable completely up to you
[ 0 ]
It only needs a union if the `ConstructEntityQueryDesc` does NOT require auth. Its weird, I know, the transform sync needs to have an overhaul.
[ 1 ]
if the function fails in the middle this will not get executed and the connection will be left open and you will not be able to drop a db etc
[ 0 ]
WooCommerce still not support at all the ability to don't apply VAT for UK if you're an EU store, this is almost ridiculous. And I'm even not sure why there is a Standard TAX rate table taht you can manually edit if it will be ignored during checkout.
[ 1 ]
i think we should drop the hardware from option name
[ 0 ]
what's with this ugly cast?
[ 1 ]
stupid me
[ 0 ]
not trove as in the openstack project trove but rather the distutils trove classifiers ref below yeah i agree it is misleading ref distutils trove classification
[ 0 ]
bool returns false why do not you just say self ecomnoverification false
[ 0 ]
good catch re why we were emitting the events in the wrong order previously this change happens to fix that too the dns create create event should come before dns recordset create etc
[ 0 ]
> Please add `Unfavorited` to `/cspell-dim-dict.txt` Done, I spent more time than I like to admit debating between "Unfavored" & "Unfavorited". It's kinda clumsy \U0001f605
[ 1 ]
Typo "appologies". Also "litte time" a bit further down. The "I see battle-sign" thing is also rather weird. It sounds more like something an ogre or saurian might say.
[ 1 ]
gross, but I will take it
[ 1 ]
I want to make this configurable. But damn, I wanna see this in the headset!
[ 1 ]
is there any reason you renamed unthrottleable to unpausable i think that this queue is unthrottleable
[ 0 ]
i think we want to get away from using bt hdr for this kind of thing layer specific is bad hack that is not great for readability
[ 0 ]
nit nullptr stream override params
[ 0 ]
extensions browser extension prefs cc error no matching function for call to makecheckopvaluestring dcheck eq observer list begin observer list end base logging h note expanded from macro dcheck eq define dcheck eq val val dcheck op eq val val base logging h note expanded from macro dcheck op eat stream parameters logging makecheckopvaluestring etc etc etc c template nonsense
[ 0 ]
Fuck actually honestly I have no real idea what'd fix it properly. Just leave it like this would be my best guess, for these cakes it doesn't matter if it's called twice anyways.
[ 1 ]
i think its merge shit >_< w8 few minutes
[ 1 ]
These names are weird
[ 1 ]
ALSO just fucking use `this || that` lmao but not as a define that's dumb
[ 1 ]
what the fuck is this, shouldn't this be handled by the staff?
[ 1 ]
jni entrypoints are not used by quick remove them it is also wrong to do this direct trampoline with them as they may throw exceptions
[ 0 ]
in the local settings py file with
[ 0 ]
Yet another Microsoft disgrace: they make so many billions from us customers that it's hard to count, but such simple and obvious feature requests are ignored for years.
[ 1 ]
it is ridiculous that the tests care
[ 1 ]
how about get rid of the arg i e view structure emplace back
[ 0 ]
Oh damn, that does seem to be the intent. That sucks. Disregard my comment here.
[ 1 ]
it will be run undef haproxy
[ 0 ]
only if i add some kind of helper method to do auto conversion to bool which would have the nasty side effect of allowing the enum values to be auto cast to int which kind of reduce the value of enum class i t o type checking
[ 0 ]
Are we going to list all possible interactions here? (like, what if there is a pushable on the matrix border?) Sounds like a potential maintenance hell
[ 1 ]
why a new function see below
[ 0 ]
this backslash is unnecessary and should be removed
[ 0 ]
kill those sections below if they are not needed
[ 0 ]
I is semi standard, and I think it's fine enough here ~~says the man who wrote this shitcode fuck I need to touch this up~~
[ 1 ]
it should be versioned as in methods below
[ 0 ]
Ah fuck me thought I fixed that, was tired when I made this pr so its not suprising theres some sloppy mistakes.
[ 1 ]
that's the shittiest name ever, that's idle frames :S
[ 1 ]
i think we should not blindly assume it x atleast we should grep for xeon
[ 0 ]
Wacky. Thanks, @drodriguez. @swift-ci create
[ 1 ]
Yeah, seems reasonable to fix the behavior so the config hierarchy gets empty sections so you don't have to do wacky stuff like add random junk to your sections for them to exist
[ 1 ]
Indentation is screwed up.
[ 1 ]
those lines glbindtexture and glteximage d are useless since your textures are created in texturetest cc by extension you can remove one of the glgeterror calls
[ 0 ]
isn t the above line equivalent to candidates set new settings since if unknown keys contained anything then an exception would have been thrown above
[ 0 ]
> Guys i just found out the solution. you only need to reinstall the previous version blockchain. That is just ridiculous bug of the shitty new version..... What version did you install to fix the problem?
[ 1 ]
This log message is a bit strange. I would do "... size={}" or something.
[ 1 ]
kill the extra spaces
[ 0 ]
super nit space after neutronworkers
[ 0 ]
also for reference some of the chemistry procs for add/remove and shit become pretty costly over time and like 75% of atoms have reagent holders so this helps i swear
[ 1 ]
nearly every fucking use of this proc involved spawn(0)
[ 1 ]
Because it looks like it literally couldn't matter fucking less to me http://puu.sh/plOcc/0bb0432dce.png
[ 1 ]
please move the quotes to the configfile so that is as valid as possible this way only the litteral values need to be replaced which is cleaner
[ 0 ]
i personally like tests that also execute all the methods
[ 0 ]
it's for player drakes, and 2 seconds is absolutely bonkers.
[ 1 ]
Ah, damn it, i thought this was a neat (and time saving) way to prevent libs from being added several times without checking every time.... well i guess we need to check for it then...
[ 1 ]
ah right shit I forgot. nvm
[ 1 ]
what is about real path readlink f
[ 0 ]
i would remove either mbufferwidth or mbuffersize leaning towards removing mbufferwidth as width sounds weird to me for this purpose
[ 0 ]
shit. right. yeah. I'll fix that or add some other words.
[ 1 ]
rebase failure here will fix
[ 0 ]
this guy here he has an rabbitmq com email address says you have to set them i also tested without these parameters set and it does not work correctly
[ 0 ]
you really do not need to derive from std less if anything you should be deriving from std binary function but it is really not needed for std set to work correctly so i would just trash that base class
[ 0 ]
agreed i think that the algorithm should be record metadata dirty state before doing anything remove the key from instance metadata if metadata was not dirty before it should not be dirty after this since the changes have been synced if it was then it was not because of this and it should remain dirty thoughts
[ 0 ]
This is a sweet mother fucking way of using it! I am sold!
[ 1 ]
ahh bylat, what the hell is this, just keep a global ticket counter
[ 1 ]
you could try state sending cmd state waiting cmd done
[ 0 ]
Actually fuck it I'll just update the maps properly in like 2 days
[ 1 ]
Damn! Apparently I can't read code anymore :(
[ 1 ]
Description is awful. Fix that.
[ 1 ]
Sneaky bastards... They should be quartered.
[ 1 ]
and this line should now be removed
[ 0 ]
shouldn t need this mock since there is a fixture that mocks it instead try self apt read return value resp
[ 0 ]
inline this empty destructor
[ 0 ]
Had my first attempt at the lobby chrome. Created a total abomination, viewable at the usual https://github.com/UberWaffe/OpenRA/tree/ProductionModifier. Will attempt surgery tomorrow.
[ 1 ]
Yeah that's kind of ridiculous. Tbh our efforts are better spent knocking down the existing style errors and warns to zero and then autofail PRs that introduce new ones. Rather than coming up with new unsolved tests.
[ 1 ]
cannot this be done as a decorator still it should be executed on call see
[ 0 ]
hmm would it be more clear if we renamed these decorators async passthru and sync passthru
[ 0 ]
Are there any plans to fix this issue? It's really annoying and ridiculous that a framework mature like Xamarin.Forms has bugs like this. The CollectionView is one of the most basic UI components of the most apps.
[ 1 ]
damn it, i thought I got them all...
[ 1 ]
may i suggest solum common yamlutils for consistent behavior wrt load and dump
[ 0 ]
this id will not be correct for anyone other than the original author so should be removed
[ 0 ]
we do not have the u of crud here
[ 0 ]
i do not need sudo anymore for testing with the dummy daemon inherit from the base test case instead of sudotestcase
[ 0 ]
elliott do you remember why we need to wait for threads to stop here daemons stop already waits until all daemons are stopped by calling thread join on all of them
[ 0 ]
this is a buggy line you are registered things with the seed but we need an end to end test here so you should not do this they were registered with the undercloud by undercloud sh
[ 0 ]
Oh crap, i remvoed the wrong thing, damn it
[ 1 ]
this init is redundant
[ 0 ]
nit indentation imo it is better to put all arguments on the new line ml db add network segment self ctx session network id segment is dynamic is seg dynamic same on line
[ 0 ]
apologies if this is another silly observation the pipeline has the proxy logging filter twice what s the reason for that
[ 0 ]
this seems like an upstream bug is it reported sane defaults win
[ 0 ]