[Bug 953514] Review Request: varnish-vmod-querystring - QueryString VMOD for Varnish

[bugzilla at redhat.com]

https://bugzilla.redhat.com/show_bug.cgi?id=953514



--- Comment #11 from Dridi Boukelmoune <dridi.boukelmoune at gmail.com> ---
(In reply to Ingvar Hagelund from comment #10)
> I haven't looked in detail on this yet, but if I understand you correctly,
> you have to get the source tree of varnish (like the .src.rpm, for example)
> to be able to build the vmod. Is this correct?

Almost correct, this is currently just another "upstream":
# VMODs need a varnish build from the source, this is by design
Source0: http://github.com/Dridi/%{VMODNAME}/archive/v%{version}.tar.gz
Source1: http://repo.varnish-cache.org/source/varnish-%{VARNISHVER}.tar.gz

> There are several ways to accomplish this. One is to create a package
> varnish-devel, containing the complete source tree, including fedora
> patches. That is the easiest way out, I guess.

This would probably help, but where would it land in the filesystem ? Something
like "/usr/src/varnish" ?

My concern is that it would be a read-only source tree, which would not be
enough. If you don't compile varnish and just use C headers, and python scripts
for vmod-related code generation, then you have to significantly patch the vmod
build to make it use binaries in /usr instead of those built in the source
tree.

Being both the upstream and the packager, this isn't really a problem. My plan
was to submit more useful vmods, and I don't know which solution would be
simpler.

> Mark that the vmod at least during the varnish-3.x series, has to have a
> hard dependency on the varnish version it is built against.

Yes and still true for Varnish 4, in the spec this is enforced with:
Requires:      varnish = %{VARNISHVER}

> If there is
> different versions of varnish available, for example while an update is
> rushed out, the vmod might break unless it's updated at the same time. This
> is impractical, but I can't see any other solution. Coordination could
> happen via the varnish-dist email list, I guess.

The vmod *will* break even if it builds correctly, because the ABI check is
done at runtime. So yes we would probably need to coordinate our releases. The
current version of the module works with both varnish 3.0.4 and varnish 4. The
varnish 4 support might break until it's actually released, but I'm regularly
checking the master branch.

I hope this answer your questions,
Dridi

-- 
You are receiving this mail because:
You are on the CC list for the bug.
Unsubscribe from this bug https://bugzilla.redhat.com/token.cgi?t=cqJgJWEWlt&a=cc_unsubscribe