Deprecate long and add notes on integer types

[hlandau]

Fixes https://github.com/openssl/openssl/issues/18338

[hlandau]

Updated.

[hlandau]

Updated.

[hlandau]

Updated.

[hlandau]

Updated.

[hlandau]

Updated.

[hlandau]

Updated.

[mattcaswell]

There are conflicts in this PR. Please can we get them resolved and then we can start a vote on this.

[hlandau]

Rebased.

[mattcaswell]

@hlandau - there seems to be some unresolved comments from @kroeckx. Can you respond before we take this to a vote?

[hlandau]

@mattcaswell I've responded, I don't think changes are needed.

[hlandau]

Still current, possible candidate for OTC discussion tomorrow.

[mattcaswell]

I have an open action to start a vote on this. I can start that shortly if you wish?

[t8m]

I do not see a need to discuss this again at OTC meeting. It was already discussed and there were no substantial changes since. So IMO the vote can be started.

[hlandau]

Sounds good.

[mattcaswell]

Starting the vote for this: deprecate long and add notes on integer types

[mattcaswell]

Vote: [+1]

[t-j-h]

Vote: [-1]

The recommendation on using int8_t is enough for me to vote against this suggestion. I also see no reason to avoid using long and using size_t for array indexes we know are small makes no sense.

[t8m]

Vote: [+1]

[paulidale]

Vote: [-1]

[levitte]

From the responses, it looks like this isn't really ready, and needs another round of fiddling.

Vote: [0]

[levitte]

Re size_t for all array indexes, I would say that consistency is a factor. It makes no sense to have different index types depending on the size of the array (what would be the pivot size?)

[kroeckx]

I'm voting +1

[mattcaswell]

Current voting position:

  Dmitry     [  ]
  Matt       [+1]
  Pauli      [-1]
  Tim        [-1]
  Richard    [ 0]
  Shane      [  ]
  Tomas      [+1]
  Kurt       [+1]
  Matthias   [  ]
  Nicola     [  ]

Still need votes from @beldmit, @slontis, @mspncp and @romen.

[t8m]

Re size_t for all array indexes, I would say that consistency is a factor. It makes no sense to have different index types depending on the size of the array (what would be the pivot size?)

Hmm, I actually agree with Tim that size_t should be used only in case we do not know the index will be small. Pivot size? Anything that clearly fits into 31 bits of the guaranteed int size? But I would not make it hard requirement to use int as index for these smallish arrays.

[t8m]

Anyway I still think this is an improvement for start and can be adjusted further in further PRs.

[romen]

Vote: [+0]

[beldmit]

I'd like to see some amendments according to the discussion. I strongly like the approach as a whole. [+0]

[mspncp]

Vote: [0]

[t8m]

Now it's on @slontis to determine the outcome of the vote.

[slontis]

Vote: [-1] I think it probably goes too far - it needs to be simpler to pass.

[mattcaswell]

I have now closed this vote. The final result was:

Topic: Deprecate long and add notes on integer types
Proposed by: Matt Caswell
Issue link: https://github.com/openssl/technical-policies/pull/51
Public: yes
Opened: 2022-07-25
Closed: 2022-08-02
Accepted:  no  (for: 3, against: 3, abstained: 4, not voted: 0)

  Dmitry     [+0]
  Matt       [+1]
  Pauli      [-1]
  Tim        [-1]
  Richard    [ 0]
  Shane      [-1]
  Tomas      [+1]
  Kurt       [+1]
  Matthias   [ 0]
  Nicola     [+0]

[mattcaswell]

It seems to me that there would still be some support within OTC for some variation of this PR, so I will leave this open for now to enable OTC to determine next steps.