Utility module for varnishtest

Manual section



import vtc [from "path"] ;

VOID barrier_sync(STRING addr, DURATION timeout)

BACKEND no_backend()

STEVEDORE no_stevedore()



VOID workspace_alloc(ENUM, INT size)

INT workspace_free(ENUM)

VOID workspace_snapshot(ENUM)

VOID workspace_reset(ENUM)

BOOL workspace_overflowed(ENUM)

VOID workspace_overflow(ENUM)

INT typesize(STRING)



The goal for this VMOD is to provide VCL users and VMOD authors means to test corner cases or reach certain conditions with varnishtest.

VOID barrier_sync(STRING addr, DURATION timeout=0)

When writing test cases, the most common pattern is to start a mock server instance, a Varnish instance, and spin up a mock client. Those entities run asynchronously, and others exist like background processes (process) or log readers (logexpect). While you can synchronize with individual entities and wait for their completion, you must use a barrier if you need to synchronize two or more entities, or wait until a certain point instead of completion.

Not only is it possible to synchronize between test entities, with the barrier_sync function you can even synchronize VCL code:

sub vcl_recv {
    # wait for some barrier b1 to complete

If the function fails to synchronize with the barrier for some reason, or if it reaches the optional timeout, it fails the VCL transaction.


BACKEND no_backend()

Fails at backend selection.

STEVEDORE no_stevedore()

Fails at storage selection.


It can be useful to crash the child process in order to test the robustness of a VMOD.


Block the current worker thread.


It can be useful to put a workspace in a given state when testing corner cases like resource exhaustion for a transaction, especially for VMOD development. All functions available allow to pick which workspace you need to tamper with, available values are client, backend, session and thread.

VOID workspace_alloc(ENUM, INT size)

VOID workspace_alloc(
   ENUM {client, backend, session, thread},
   INT size

Allocate and zero out memory from a workspace. A negative size will allocate as much as needed to leave that many bytes free. The actual allocation size may be higher to comply with memory alignment requirements of the CPU architecture. A failed allocation fails the transaction.

INT workspace_free(ENUM {client, backend, session, thread})

Find how much unallocated space there is left in a workspace.

VOID workspace_snapshot(ENUM)

VOID workspace_snapshot(ENUM {client, backend, session, thread})

Snapshot a workspace. Only one snapshot may be active at a time.

VOID workspace_reset(ENUM)

VOID workspace_reset(ENUM {client, backend, session, thread})

Reset to the previous snapshot of a workspace, it must be the same workspace too.

BOOL workspace_overflowed(ENUM)

BOOL workspace_overflowed(ENUM {client, backend, session, thread})

Find whether the workspace overflow mark is set or not.

VOID workspace_overflow(ENUM)

VOID workspace_overflow(ENUM {client, backend, session, thread})

Mark a workspace as overflowed.

INT typesize(STRING)

Returns the size in bytes of a collection of C-datatypes:

  • 'p': pointer

  • 'i': int

  • 'd': double

  • 'f': float

  • 'l': long

  • 's': short

  • 'z': size_t

  • 'o': off_t

  • 'j': intmax_t

This can be useful for VMOD authors in conjunction with workspace operations.