hsts: duplicate live hsts data in curl_easy_duphandle().

[whatchamacallem]

curl_easy_duphandle() now copies live HSTS entries (aquired from Strict-Transport-Security response headers) into the cloned handle. Previously only file-based entries were duplicated. Adds Curl_hsts_copy() to handle the transfer. Expired entries are skipped.

[whatchamacallem]

This looks ready to merge. I'll let you handle it from here. (I ended up rebasing the commit to merge your fix to a flakey test and a couple times more to pass the linter.)

[bagder]

How about adding a libtest that verifies this?

[vszakats]

Suggested change

	static CURLSTScode hstsread1922(CURL *curl, struct curl_hstsentry *e,
	void *userp)
	static CURLSTScode t1922_hstsread(CURL *curl, struct curl_hstsentry *e,
	void *userp)

minor indent nit, and also the naming scheme as in most other tests.

[vszakats]

Suggested change

	#if (SIZEOF_TIME_T < 5)
	#if SIZEOF_TIME_T < 5

drop parents, to make it like most of the code

[whatchamacallem]

I was in the middle to dorking with this PR when you made suggestions. I have applied all your suggestions to the latest commit.

[vszakats]

Suggested change

	struct state1922 {
	struct t1922_state {

sync with naming scheme used in most tests.

[vszakats]

Suggested change

	static CURLSTScode hstswrite1922(CURL *curl, struct curl_hstsentry *e,
	struct curl_index *i, void *userp)
	static CURLSTScode t1922_hstswrite(CURL *curl, struct curl_hstsentry *e,
	struct curl_index *i, void *userp)

fix indent, and use common naming scheme.

[vszakats]

Suggested change

drop empty line.

[vszakats]

This file needs to be added to the local Makefile.inc to add it to dist and pass CI.

[whatchamacallem]

@vszakats @bagder

I don't know what to do about this one CI issue.

https://github.com/curl/curl/actions/runs/26723735880/job/78755299214?pr=21809

Pseudo-terminal will not be allocated because stdin is not a terminal.
/bin/sh: cmake: not found
Error: Process completed with exit code 127.

[whatchamacallem]

This looks ready to merge. Let me know if you want it rebased onto master again. The midnightbsd failure was also failing on master. The other two failures look like overloaded CI servers timing out and running out of resources.

[bagder]

Shouldn't the test also use the duplicated handle to "prove" that the HSTS data is actually copied over? Otherwise the test doesn't really verify that.

[whatchamacallem]

@bagder @vszakats All tests pass. I didn't rebase because master had a failing test.

(I swear just using file globs is easier than maintaining lists of files. We have git for tracking the contents of directories. Fixing the whitespace in a list of files feels like masochism after you have experienced freedom.)

[whatchamacallem]

@bagder @vszakats Let me know if there is anything else for this HSTS patch to land. The next step would be to emulate TLS somehow, I'm hoping this is good enough.

[bagder]

Thanks!