add a flag to control the hash checking

[Trial97]

Address the compatibility issue with zlib-ng-compat, which is now shipped in place of zlib in various Linux distros. This change introduces an option to disable hash mismatches reported on processors.
I plan to integrate it with the ForgeWrapper.

[neoforged-pr-publishing]

• Publish PR to GitHub Packages

[shartte]

How realistic is shipping zlib-ng?

[Trial97]

Right now, I know that cachyos defaults to zlib-ng instead of baseline zlib.
Other than that, I do not know of other distributions that do that, but I would expect either the installer to behave the same regardless of the system libraries or have a way to bypass the check.

[shartte]

Right now, I know that cachyos defaults to zlib-ng instead of baseline zlib. Other than that, I do not know of other distributions that do that, but I would expect either the installer to behave the same regardless of the system libraries or have a way to bypass the check.

I actually mean via JNI bindings we ship ourselves to use the same zlib everywhere.

note: It is however true that we should not necessarily rely on reproducible compression to begin with, but I wonder if disabling the hash check is actually to our benefit.

[Trial97]

I want to have the users who encounter this issue have an alternative to installing the loader.
Yeah maybe it is not the best alternative, maybe there is a way to set some settings for compression that would ensure reproducibility on all cases. But that part is to advanced for me.
Here is the original discussion on Neoforge Discord: https://discord.com/channels/313125603924639766/915304642668290119/1346007687221219369

[shartte]

I want to have the users who encounter this issue have an alternative to installing the loader. Yeah maybe it is not the best alternative, maybe there is a way to set some settings for compression that would ensure reproducibility on all cases. But that part is to advanced for me. Here is the original discussion on Neoforge Discord: https://discord.com/channels/313125603924639766/915304642668290119/1346007687221219369

While doing a reference check might work I am worried that zlib-ng only compresses differently in some, but not all cases. Then a reference check might look like the zlib implementation used by the JDK is what we expect, but it still produces differences in some cases.

Regarding your change: I'd not merge this flag as-is, instead I'd want the flag to continue to print that there was a hash mismatch, but simply not delete the resulting file (and continue).

[shartte]

I'll try to take a closer look tomorrow

[neoforged-releases]

🚀 This PR has been released as LegacyInstaller version 3.0.39.

[Trial97]

Thanks for merging this.
Are there any additional steps for the newer version to use this installer version?
And I presume there will be no regeneration of old files, right?

[shartte]

Thanks for merging this. Are there any additional steps for the newer version to use this installer version? And I presume there will be no regeneration of old files, right?

Usually not, no. But NF needs to bump the installer version to make use of this (gradle.properties over there).

edit: we have backported a shit ton to older versions so it's possible to backport this too, actually.

[SolsticeClouds]

Hi, so sorry to bother but is there any chance this could be backported to 47.1.106 or similar? i want to play a 1.20.1 modpack but i'm having this issue with the hash check due to having to use an older version of NeoForge.