Nils Goroll

Updated suggestion after bugwash: * `-s` arguments only define storages, not purposes * parameters define default storages for purposes, their default value is the first `-s` argument. For a transition...

The case here was with a vmod function returning `NULL`, but I think @gquintard 's suggestion has some merit to it: `NULL` could signify "undefined/unset", but still evaluate to `""`...

Summary of bugwash discussion: The current proposal basically boils down to * for type conversions/folding, preserve `NULL`: `Typea(NULL) => Typeb(NULL)` * for string contexts, convert `NULL` to `""`: `String(NULL) =>...

I want to prepare a writeup of the broader topic

On the Lifetime of VCL data =========================== PR #3546 fixed a special case of this problem: References to storage objects are only valid until we release the reference to the...

Re @bsdphk > we have thereby decided that arguments to `VRT_SetHdr()` cannot live on the stack. This is actually a very good example because that function's signature is ```c VRT_SetHdr(VRT_CTX...

> That would require us to wrap all the relevant VCL types in a struct vmod_priv I would think the vmod would only need to do it if it has...

This ticket has been discussed during bugwash. This note is to document what I believe to be general consensus. Please comment or leave a 👍🏽 or 👎🏽 ---- VRT and...

I do fully agree that we should make improvements in this area and I think the general approach of adding ``vcl_raw {}`` and a new "pipe-fsm" is the right one....

notes from bugwash from sejerman wis se bullat points: * concerns have been raised to complicating the `builtin.vcl`. Agreement was found that modularizing it would help. the `vcl_*` prefix was...