Deprecate the "print" functions, especially the undocumented ones.

[bob-beck]

We have a lot of SOMETHING_print and related functions in the libraries.

Many are undocumented, and of those, there's really one user (apps).

As what these print is basically a bespoke thing, and not documented in any way, this
functionality really does not belong in the library - it belongs in the thing that cares about it,
which is the app.

With possibly a few necessary exceptions (which must be documented) Rather than keep these as undocumented "can't touch this" stuff in libcrypto and libssl, these should be deprecated from the library, and moved to the apps library. This may entail ensuring that accessors for opaque types are present so that API
is available so that these functions may be implemented outside of the library, rather than needing to
call internal functions (necessitating these being in the library).

Ideally, the library probably should not have "print" functions in general.

[Sashan]

Many of those functions are wrappers around libc. the exception is support for legacy windows if I remember correctly.

[t8m]

Hmm... this sounds like a noble goal, however do we have any idea how much are these functions used outside of the OpenSSL application? I.e., in 3rd party apps? I assume there might be apps that care about textual representation of some of these objects.

[bob-beck]

most of the currently undocumented ones are used very little, to not at all (outside of the app)

some of the documented ones have outside use, and as part of this I'm gonna go peel a list out of the things which I have access to that have wide exposure to zillions of third party things.

As for things that care about a textual representation of things, this is actually a problem in many cases, it tends to fall
into a few categories:

0. It has frequent use for (something) legitimate to print out, meh, we may keep it around, and ensure use and limitations
   are documented.
1. it's debugging code and they don't care - (copy code from the app and be done with it, or keep using the deprecated funciton for a while).
2. they actually care what is printed, because god forbid - they parse it.
   this becomes a real pain if we change anything that changes printed output, because we end up with output on a tty
   becoming hyram's law, and we break things if the output format changes.
   The various X509_subject_oneline and friends are like this - so these ones must be treated with some care, but we
   really want to get out of that business.

The first targets for this are simply the stuff that is both undocumented, and has essentially no use elsewhere, then we work out way through some of the others. I don't know if it's reasonable for us to be completely print function free, but I think
starting with a lot of the 43 currently undocumented print functions in libcrypto can probably cut down that list a lot for a start.

[bob-beck]

Of note, I do not include things in the category like things which print out actually standard textual formats, like PEM and the like. We have every reason to be able to do that.

I'm only referring to things that spit out a bespoke printed version of something.