Martin Blix Grydeland

The reason that the ban mutex is being held across the STV_Ban* calls aren't to protect the ban data structures from concurrent access (which as I read it the commit...

The important part of the contract between stevedore and varnish here is that `STV_BanInfoDrop()` is called in the exact same FIFO order with the same ban information as the `STV_BanInfoAdd()`...

While working on #3480, I had a quick look at this. I believe there are cases where this patch set as is fails and causes deadlocks. Specifically, it assumes that...

> we should add to the release notes procedure that all the steps up to `make dist` should be made on a system with latest autotools packages. I'm uncertain what...

I believe this is the result of too agressive write vector caching before doing the first writev() system call against the backend. While piecing together the writes that make up...

I believe the events @nigoroll describes in https://github.com/varnishcache/varnish-cache/issues/3775#issuecomment-1020029452 is a plausible explanation for the behaviour described here. It should be noted that the parent child vtx relationships in the log...

I guess the discrepancy to investigate is that OF_IMSCAND doesn't track whether we sent IMS/INM on the fetch, only if we could've done that. And especially for zero-length objects, it...