Closed nigoroll closed 6 years ago
I'll check that this change doesn't break older releases. I expect not, I already performed similar checks for vmod-querystring.
Thanks!
FYI travis ci failed because we are installing Varnish from repo.v-c.org...
Have pushed more changes I guess a 5.2 branch before merge would be warranted
@nigoroll we aren't retiring vmod-softpurge, please drop 1e2e8b7. We will keep compiling it for Varnish 4.1, I'll take care of that.
@Dridi feel free to cherry-pick what you like.
But are you sure keeping a single branch makes sense? My impression is that we've reached a point where we'd just have too much #ifdef
ery
To fully support [4.1-master] I need this patch:
diff --git i/src/vmod_header.c w/src/vmod_header.c
index ee0b02a..6bb26f9 100644
--- i/src/vmod_header.c
+++ w/src/vmod_header.c
@@ -27,12 +27,19 @@
*/
#include "config.h"
+
#include <stdlib.h>
#include <string.h>
#include <pthread.h>
#include "cache/cache.h"
+#include "vdef.h"
+
+#define VCL_RET_MAX 1
+enum vcl_event_e;
#include "vrt.h"
+#undef VCL_RET_MAX
+
#include "vcl.h"
#include "vcc_header_if.h"
diff --git i/src/vmod_vsthrottle.c w/src/vmod_vsthrottle.c
index cae5865..1b6d2a5 100644
--- i/src/vmod_vsthrottle.c
+++ w/src/vmod_vsthrottle.c
@@ -36,8 +36,14 @@
#include <errno.h>
#include "vdef.h"
+
+#define VCL_RET_MAX 1
+enum vcl_event_e;
#include "vrt.h"
+#undef VCL_RET_MAX
+
#include "vcl.h"
+
#include "vas.h"
#include "vtim.h"
#include "miniobj.h"
diff --git i/src/vmod_xkey.c w/src/vmod_xkey.c
index 4355c52..d192443 100644
--- i/src/vmod_xkey.c
+++ w/src/vmod_xkey.c
@@ -33,7 +33,13 @@
#include <string.h>
#include "cache/cache.h"
+#include "vdef.h"
+
+#define VCL_RET_MAX 1
+enum vcl_event_e;
#include "vrt.h"
+#undef VCL_RET_MAX
+
#include "vcl.h"
#include "vsha256.h"
I'm not too happy about the vdef,vrt,vcl ordering because past Varnish versions forced us to pick vcl,vrt in order to make the event signature available to VMODs. We'll need to discuss this further, I'll take care of vmod-softpurge independently.
We had one more change in varnish-trunk regarding #include
order and rules. We'll revisit this after the 3 months milestone and decide then how we manage compatibility with the different Varnish branches.
https://github.com/varnishcache/varnish-cache/commit/b6921558b9e5ecb431524f9d5dc9aea5af7bd24b
force-pushed an update to get ahead in the meantime. the travis-failure is unrelated.
As I said before, we will wait until Dec 15th ABI stabilization to revisit this, but I don't think we need the VSHA256 change:
https://github.com/varnishcache/varnish-cache/commit/8844d6ba3a0ef8e984c6ac70a3acf2cf93f542f2
required to build against current master