docs(README): add note about hash collision problem

[user1736]

What kind of change does this PR introduce?
Added documentation about hash collision problem described here

Did you add tests for your changes?
Not applicable

If relevant, did you update the README?
Yes

Summary
I decided to document workaround from discussion in #413 to make people more aware of that problem

Does this PR introduce a breaking change?
No

[jsf-clabot]

All committers have signed the CLA.

[user1736]

Is anybody alive there? It'll be nice to have it triaged at least

[alexander-akait]

@user1736 can you provide example with hash collision?

[alexander-akait]

Maybe we can fix it on our side without documentation

[user1736]

@evilebottnawi did you check the conversation in #413? This is what I'm talking about

[alexander-akait]

@user1736 yes, not reproducible test repo, if you provide we can fix it in next major release (will be released in near future)

[user1736]

@evilebottnawi, sorry, but did I miss something? #413 (comment) you said that documentation is welcome and the issue labeled with doc tag.

If documentation not needed anymore feel free to close this PR. I'll consider creating test repo for that problem then.

[alexander-akait]

@user1736 we should first look on reproducible repo, if it is won't fix we documented this

[ScriptedAlchemy]

Agreed, Ive seen collisions before. This was back in Webpack 1. I believe someone on my team pached this waaaay back.

We need to reproduce this with a repo to help. If you are reproducing it, then just throw us the webpack file and a hello world with this bug.

Could it be related to using contenthash bs chunkhash or something? I know some hashing is based on bytes and can lead to collisions. I like to use the BEM style path along with a hash on the end.

Really, it comes down to the being able to reproduce the issue. It’s in general, our main struggle - we cannot catch race conditions till someone’s give them to us. There’s no SDETs in OSS.

If there is indeed an issue, it should be documented - but the webpack team is rightfully hesitant to merge a PR which documents a potentially diasterous issue when there so many variables in modern Webpack configurations. Let’s not set the alarm bells of too soon 😄

[alexander-akait]

@ScriptedAlchemy because we use salt and very long key for hashing and problem with collusion right now should don't exists

[ScriptedAlchemy]

@evilebottnawi Agreed, I watched your team move to the better hashing / my team was involved a long time ago in solving similar issues in webpack hashing.

There should not be any collisions based on my experience with webpack hashing functions.

Again, we need a repo to help @user1736 and are more than willing to do so. But we must understand and see the issue to trace the problem.

[alexander-akait]

Can't reproduce so it is not actually, anyway feel free to open an issue/a PR if you faced with bug