- 23 Feb, 2016 15 commits
-
-
Martin Blix Grydeland authored
n_objectcore is incremented by ObjNew(), so HSH_Insert() should not increase it again.
-
Martin Blix Grydeland authored
HSH_Insert() needs to take the same precautions as HSH_Unbusy() in order to not race against e.g. the ban lurker. In order to do this, we pass it the ban that the objcore should be associated with, and it will take care of calling BAN_RefBan() correctly.
-
Martin Blix Grydeland authored
BAN_FindBan looks up a ban based on a ban timestamp and returns a pointer to the ban. Returns NULL if no ban with that timestamp is present. BAN_RefBan takes an objcore and a ban. It grabs a reference to the ban and associates the objcore with that ban.
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
Fixed size attributes has it's space always reserved. Variable size attributes change in size, and their combined total length needs to be given to allocobj. Auxiliary attributes do not need to be predeclared. The code for setting and getting the attributes is generated from a table that lists the different attributes according to their type. Add documentation about the behaviour of the get/set functions.
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
This is to make it clear that one can not change an object attribute through the returned pointer from ObjGetattr()
-
Martin Blix Grydeland authored
The objcore won't gain a boc in a way that would call for locking the OH when checking if it wasn't there in the first place. So if the oc->boc is NULL, return early. This saves a mutex lock/unlock in the common case.
-
Poul-Henning Kamp authored
only once, once we know it.
-
Poul-Henning Kamp authored
This allows Martin to stash attributes on going into BOS_STREAM.
-
Dridi Boukelmoune authored
-
Poul-Henning Kamp authored
a request for zero bytes. Suggested by: github::onovy
-
Poul-Henning Kamp authored
-
Federico G. Schwindt authored
This has been available via the std module for some time now. OK'd by phk@.
-
- 22 Feb, 2016 8 commits
-
-
Dridi Boukelmoune authored
-
Dridi Boukelmoune authored
Autotools stuff is kept out of the picture, except pkg-config(1) which is essential for testing foreign code against a fresh Varnish build.
-
Dridi Boukelmoune authored
Parameters are not checked and passed as is to varnishd command lines. The parameters are prepended at the beginning of the command line to preserve parameters needed by varnishtest or used in test cases.
-
Dridi Boukelmoune authored
-
Dridi Boukelmoune authored
-
Dridi Boukelmoune authored
The macros are still here but ${vmod_std} is now expanded to std and so on.
-
Dridi Boukelmoune authored
Instead, varnishtest users should set up a proper PATH, just like `varnishtest -i` does to use binaries from the build tree.
-
Nils Goroll authored
-
- 20 Feb, 2016 1 commit
-
-
Poul-Henning Kamp authored
-
- 19 Feb, 2016 3 commits
-
-
Federico G. Schwindt authored
-
Federico G. Schwindt authored
The former are kept for backward compatibility as aliases for now. While here update some related doc and remove dead code.
-
Federico G. Schwindt authored
-
- 18 Feb, 2016 2 commits
-
-
Guillaume Quintard authored
Thanks fgs for catching it
-
Dridi Boukelmoune authored
-
- 17 Feb, 2016 7 commits
-
-
Guillaume Quintard authored
-
Guillaume Quintard authored
-
Devon H. O'Dell authored
Modern versions of gcc and clang support various runtime sanitizers that can be used to find nasty bugs at runtime. Note that not all of these sanitizers may be active at the same time. However, these tools are aimed at developers and likely require at least some level of reference during testing while enabled -- which discouraged me from doing a matrix of conditionals in autohell.
-
Dridi Boukelmoune authored
-
Nils Goroll authored
-
Nils Goroll authored
We'll need to change a couple of signatures in the call paths getting here to propagate up the failure.
-
Nils Goroll authored
This fix avoids a WS_Alloc panic when pushing fetch processors and should thus avoid panics by overflowing the backend workspace in vcl. This overflow is logged as FetchError "Bo workspace overflowed". Other panic points due to workspace_backend being set too low still exist. Sizing estimate for workspace_backend: sizeof(struct busyobj) + 3 * HTTP_estimate(cache_param->http_max_hdr) + cache_param->vsl_buffer + cache_param->http_resp_size + fetch processor memory + space required in VCL Fixes #1739
-
- 16 Feb, 2016 4 commits
-
-
Federico G. Schwindt authored
As discussed on irc. OK'd by phk@.
-
Federico G. Schwindt authored
OK'd by phk@.
-
Federico G. Schwindt authored
-
Federico G. Schwindt authored
-