Nils Goroll
Nils Goroll
FTR: the test case in question does not exercise h2, so any h2 concurrency issues should not be related to the cause
Triggered by bugwash, I think I found the root cause in pESI: https://gitlab.com/uplex/varnish/libvdp-pesi/-/commit/32b2be11f3696f9c33908c89d990e16e4f99e135
bugwash: we tend towards changing interfaces returning directors to always return with another reference held. phk said he wanted to work on this.
In the context of https://github.com/nigoroll/libvmod-dynamic/issues/117 I ran into this issue again. Here's a trivial way to trigger the issue: * add this delay to varnish: ```diff diff --git a/bin/varnishd/cache/cache_vrt.c b/bin/varnishd/cache/cache_vrt.c...
For vmod_dynamic, this is now [worked around by always keeping another reference for the duration of the task](https://github.com/nigoroll/libvmod-dynamic/commit/d8eea028563837e74b8d7f1467c0a8118f6cfc6d).
quick bugwash notes: * generally in favor * let's not break `vcl.load` * but rather add `vcl.load_files [-s state] [-p path] ...` * `-p` should be forbidden if `vcl_path` is...
bugwash: @bsdphk wants to finish a review of the CLI privilege design before making a decision on this ticket.
I would think we should remove the port from Host and fail for IP addresses.
notes from bugwash * (15:06:02) phk: so maybe first step is to formulate that strategy and put it in the man-page ? * (15:04:12) dridi: we should proceed carefully *...
I have not reviewed the formats in detail, but I agree to the overall concept _first for "receive" and last for "send"_.