While hanging, it's consuming more and more memory, until all memory (251 GiB + 148 GiB swap) is exhausted and the OOM-killer kicks in.
I've dug into this issue and found the following. When passed a particular incorrect format string (I have not tested if it happens for other incorrect format strings) the chrono function "StrftimeItems" in chrono 0.4.41 returns an apparently never-ending iterator of errors. tera tries to "collect" this never-ending iterator into a Vec. The result is the Vec produces a series of ever-larger memory allocations until either a memory allocation fails or the program is hit by the OOM killer. I'd argue there are two bugs here. On the one hand Chrono probably should not be returning a never-ending iterator. On the other hand there is no reason for the check in tera not to stop at the first error. I've filed upstream bugs with both chrono and tera, and uploaded a patched version of tera to Debian. My proposal for next steps would be. 1. Get rust-tera 1.20.0-2 into testing (with force if needed). 2. Remove the blacklist of rust-tera's autopkgtest. 3. Downgrade this bug to "important" and assign it to rust-chrono 4. Wait for a response from rust-chrono upstream to decide if we need to do anything further.
