Jeffrey Kog
- jeffreykog
- jeffreykog
- Europe/Stockholm
- Yes
- No
MC-1628![[Mojang] Nathan Adams](../../../data/assets/gravatar/dinnerbone.png){kind=small}
MCL-2138
After switching between worlds like 25 times, minecraft runs out of memory
What I expected to happen was...:
I expected to be able to switch between singleplayer worlds as much as i wantWhat
actuallyhappenedwas...:
Minecraft starts using more and more ram, making it crash because it's out of memorySteps to Reproduce:
1. Leave current world
2. Join another world
3. Leave this world
4. Join another world
And so on like 25-35 times until minecraft crashesAfter switching between worlds like 25 times, minecraft runs out of memory
I created 2 new superflat worlds with the "Redstone Ready" template.
Each time i switch to another one of these, the game consumes 60 - 100 mb more ram!What I expected to happen was...:
I expected to be able to switch between singleplayer worlds as much as i wantWhat actually happened was...:
Minecraft starts using more and more ram, making it crash because it's out of memorySteps to Reproduce:
1. Leave current world
2. Join another world
3. Leave this world
4. Join another world
And so on like 25-35 times until minecraft crashes
The launcher automatically redownloads any library from the mojang servers when the server-side checksum (.sha1) does not equal the client-side one (.sha). However, libraries with a custom maven url are not redownloaded when the checksum doesn't match.
Is this actually a launcher bug or is this intended behaviour, and when it is intended, i'm interested to know why it is intended.
~ Jeffrey Kog
Update
After looking in the source of the launcher (Yes, i know you don't want us decompiling stuff, but i only looked for the potential cause) i found out that this has to be intended, because in net.minecraft.launcher.versions.CompleteVersion in the method getRequiredDownloadables i found this:if ((!local.isFile()) || (!library.hasCustomUrl())) { neededFiles.add(new ChecksummedDownloadable(proxy, url, local, ignoreLocalFiles)); }That means that it only checks the remote checksum when the library doesn't exist locally, or has no custom maven url.
I'm interested to hear what the reason is for disabling this, because it makes downloading libraries for mods (and automatically updating them without modifying the client-side profile) a lot harder.
Have a nice day!
With the vanilla smp server 1.4.2 and minecraft client 1.4.2, no crashes for me.
This is intended. If the potion splashes right in your face you WILL get the 3:00. The more your face is away from the potion splash, the less time the potion effects will stay active
Yeah, this is intended, half slabs are still a full block, so that's why cakes just look floating
Yeah for me it's not too big of a problem that i have to restart minecraft after some switching, but it should not be necessary. Also it are just normal flat sandstone worlds with nothing special in them
Some new information about this bug.
I created 2 new superflat worlds with the "Redstone Ready" template.
Each time i switch to another one of these, the game consumes 60 - 100 mb more ram!
This is more than i expected to be.
Confirmed. Ghasts in spawners are rendered so big, that you can't see the spawner.
Can confirm this. Maybe this is because a door is a 2 block structure, however, that shouldn't matter
Also i can confirm this. I think this is because minecraft only renders a chunk when there are block changes in that chunk
Can confirm this bug. I've seen this happen before 1.3
I had the same issue on my mac. Not only in browser but also on the regular client. The problem is that you're using java 1.7
Downgrade to java 1.6 and minecraft will just run fine
I think i know what's going on here. there are 2 versions of each leaf block, the decaying one and the not-decaying one. using silktouch you pick up the decaying blocks (other metadata values).
also confirmed
Here are 2 screenshots, the first is in 1.4.7, the second in 13w02a
Same mac version, no crash. Does minecraft have permissions to the folder its running in?
I tested this again yesterday. The ram gain between world switches is smaller than it was before, but this is still an issue in the current version (1.5)
But if all these additional resources are unneccesary, then the garbagecollector should get rid of them. Right?
It did crash with that nice out of memory screen. I threw away the report, will try to reproduce it today and i will post the report here.
Also, i am not using any texturepack. Just vanilla.