remove OSSL_HTTP ->OCSP renaming#13671
remove OSSL_HTTP ->OCSP renaming#13671richsalz wants to merge 4 commits intoopenssl:masterfrom richsalz:rm-http-ocsp-naming
Conversation
Based on comments from David von Oheimb.
The OCSP_xxx names were exported anyway. Remove the define's that had the internal implementation to use new-style names.
|
Also, a comment "this function isn't used" was wrong. |
|
CI failure is the encode/decode thing and not relevant. |
There was a problem hiding this comment.
The 'solution' taken here to revert the name changes I did from OCSP_ to OSSL_HTTP_ back to OCSP is IMO entirely wrong!
Didn't you see that the data structure and the functions have been generalized from an OCSP-biased bare-bones HTTP client to a generic one where OCSP is just an instance?
I deliberately had kept the old names internally to simplify backward compatibility, though I admit that this can be confusing. Maybe a better way of achieving that would have been to rename the original function and provide #define's for providing access via the old names.
|
Sorry @DDvO I strongly disagree. Renaming the implementation but nothing else is really wrong IMO. But if the project rejects this, that's okay. I don't know the mechanism if contributors can put a hold, but perhaps some group in the project (otc or committers) wishes to discuss it. |
@richsalz, have you understood what |
|
Oops, I meant the HTTP client as introduced PR #10667. |
|
You are right, I didn't realize that |
|
All right. Let's first clarify how to solve the confusion the right way. How about the idea I gave above?
If you agree, would be nice if you can take over this, which could be done even as new version of this PR. |
DDvO
changed the title
|
I agree and will do the deprecation as you describe. |
|
Closing in favor of #13742. |
3 participants
This is built on #13620 but could probably be split.
The way things were renamed to look like
HTTP_OCSP_xxxas opposed toOCSP_xxxis wrong. The only thing enabled was that the implementation in crypto/http/http_client.c had "new" names, but a set of defines in the internal local header file crypto/http/http_local.h mapped those new names to "old" style names. The "old" names is what were exported.I spent more than an hour trying to find definitions of documented functions, etc., but failed because of the "new" names.