Fix ASCII dependency in strcpy_url and strlen_url #2535
Conversation
Commit 3c630f9 partially reverted the changes from commit dd7521b because of the problem that strcpy_url() was modified unilaterally without also modifying strlen_url(). As a consequence strcpy_url() was again depending on ASCII encoding. This change fixes strlen_url() and strcpy_url() in parallel to use a common host-encoding independent criterion for deciding whether an URL character must be %-escaped.
|
This changes behavior for redirects to URLs with "control codes" in the string, which I think is probably fine, but we need a test case or two that verify this so that we can trust the code a bit more. |
The current pull request does not change the behavior regarding control codes. Please check the condition in function To actually include control codes in the set of characters to be escaped I can simplify that condition back to For writing new test cases I need some guidance please. The test case test1138 apparently is the one that tests URLs with characters beyond 0x80. Can that be extended, or is a new test case in a new file required? The documentation in the
But the file test1138 directly contains UTF-8. Do I understand it correctly that I can directly enter the control characters in the file then? |
You're right and I was wrong. This should maintain the same functionality on ASCII systems and yet enable non-ascii systems to do the right thing. The test file format is actually totally encoding agnostic and can contain whatever is suitable so test 1138 has some UTF-8 sequences in there to verify exactly this sort of %-encoding. This test continues to work after your patch, as the CI builds and tests here already show. I will merge this asap. |
|
Thanks a lot! |
Commit 3c630f9 partially reverted
the changes from commit dd7521b because of
the problem that strcpy_url() was modified unilaterally without also
modifying strlen_url(). As a consequence strcpy_url() was again depending
on ASCII encoding.
This change fixes strlen_url() and strcpy_url() in parallel to use a
common host-encoding independent criterion for deciding whether an URL
character must be %-escaped.