shared: Introduce max_message_size_… features #331
Conversation
chrysn
force-pushed
the
size-granularity
branch
from
bac7af7 to
7f8e894
Compare
|
Thank you for the PR. Looking into why CI fails. |
|
Unlike on #333 I did not rebase this after #339 to fix CI, but instead merged main back: This enables this PR to stay on 0.7.2, i.e. to be a single fixing change on the latest release (which is how I use it on Ariel OS). @geonnave, is there anything more you'd like to review? Can we merge this as is, or should we trick around (possibly by briefly disabling CI checks) to preserve both a "feature branch simply merges back" shape and the "this can be picked into a 0.7 branch" property? (I think it's fine to merge). |
|
Aaaargh: cbindgen doesn't support switching by cfg (while it supported the old idiom of cfg'd items). [edit:] Using cfg-if doesn't help either. Emulating the old idiom doesn't scale well (the entry for the shortest size would need to be not(any(feature = "first", feature = "second", ...))), I'll look for alternatives. (ceterum censeo: exposing the API to C is a liability for this project) |
With the recently increased buffer sizes, the original test failed to achieve its goal of producing an overflow error.
chrysn
force-pushed
the
size-granularity
branch
from
d440e49 to
45b25b7
Compare
When this test failed, the resulting "Type state mismatch" runtime error was confusing -- and while currently we raise the buffer error first, there's no guarantee that that couldn't flip around. Therefore, the example is made more realistic to ensure that it's really the buffer error that triggers first.
chrysn
force-pushed
the
size-granularity
branch
2 times, most recently
from
0268159 to
c024fce
Compare
chrysn
force-pushed
the
size-granularity
branch
from
c024fce to
8f2fc25
Compare
|
Merging!! |
Closes: #330