Fix improper detection and handling of html_safe buffer in CacheHelper #2080
Conversation
|
We need a test case in order to merge this. |
|
I think test_fragment_caching in caching_tests.rb already tests the bahavior of this function. In addition, even larger changes to this function, such as 114b5e4, have been accepted without specific tests. Anyways, I'll see if I can write a test case over the weekend. |
|
@lhahne the commit you're pointing fixes some previously committed failing tests. |
|
Ok. I know a way to break the old code so I can write some tests once I discover how to change the buffer in tests. |
The output_buffer returned by CacheHelper should be html_safe if the original buffer is html_safe.
…is of the same type as the original
|
Ok. I added two tests one of which fails against the current 3-0-stable. After my fix, both tests pass. In addition, I modified CacheBuffer so that any possible new output_buffer is of the original type and made the tests reflect this. |
|
This seems to be related to #2150 |
|
Closed in PR 2219 |
|
We need to backport PR 2219 to 3-0-stable |
|
@spastorino in #2219 the regression fixed there is introduced by 03d01ec, which is not in 3-0-stable. So merging this pull request will suffice (assuming we're not backporting 03d01ec to 3-0-stable as well). Though I think we should update the cache helper test to match the test from master. |
Fix improper detection and handling of html_safe buffer in CacheHelper
This fixes the latest incarnation of #1537 where CacheHelper might try to use splice! for a html_safe buffer which converts the buffer unsafe. This change has been tested against Rails's tests and slim-lang which originally raised this issue.