- 15 May, 2013 23 commits
-
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
Remove automagic reopening logic from vsl Make vsl use the new VSL_head structure Remove most of the old functionality
-
Martin Blix Grydeland authored
into the log.
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
Show UNSEEN in bottom bar when not hiding unseened counters
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
Add a VSC_Mgt() function to get direct access to the management counters Add VSC_MgtValid(), VSC_MainValid(), VSC_IterValid() functions that call VSM_StillValid on the fantom used to produce them. Use the iter fantom when making the pointer lists, so VSM_IterValid() returns the right result.
-
Martin Blix Grydeland authored
Clear the thread specific log pointer on log close Reenable building of varnishtest
-
Martin Blix Grydeland authored
VSC type defined in vsc_types.h Rearrange the VSC types according to relevance
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
Add global variables for the type name strings. Compile in static VSC type descriptions in libvarnishapi
-
Martin Blix Grydeland authored
Don't do automagic reopening in VSC Rename vf variables to fantom Keep a list of fantoms for the points, and point the points to these. This gets rid of the fantom struct on each point, and allows for StillValid checking on groups at a time instead for each point. Rename point's class to type to be consistent with VSM terminology More class->type renames Drop the type and ident pointers from VSC_point, and use the fantom's pointers instead. Reenable building of varnishstat Don't assert on VSM file closed in VSC_Iter. It's safe to proceed anyways and will produce a zero result in that case. This makes the function consistent with how the other functions behave.
-
Martin Blix Grydeland authored
Rename VSM_FOREACH_SAFE as it isn't really safe in the vqueue sense, and removing chunks from the VSM from the client library isn't applicable anyway. Make the VSM chunk marker sizes defines. Check for shm open in VSM__itern Revisit logic of VSM__itern: Make sure that the VSM_fantom is either changed to the next, or not changed at all. Add comments. Keep copies of class, type and ident in VSM_fantom, and use these when revalidating fantoms. Do a proper VSM_Open in varnishstat Add a VSM_ResetError function Differentiate between missing HEAD marker and zero alloc sequence. The latter is an indication Varnish isn't running, the former a corrupt file. Fix newline on diag messages for consistency. Fix return value check inversion bug on stat call in VSM_Abandoned() Don't assert on VSM_Open on an already open VSM. Report success instead.
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
segment. The API will use this counter to check for abandonment. Make use of the SHM age attribute in VSM_Abandoned. This gives a less expensive way (without doing stat syscall) to check for abandonment. If there's been more than 2 seconds between calls to VSM_Abandoned and the age attribute hasn't changed in between, the stat on the file will be performed to see if the SHM file has changed. VSM_Abandoned will be the standard way to check for SHM changes in API utilities, and when it returns true, the application is supposed to call VSM_Close and then VSM_Open to reopen the SHM file. Update Makefile.am in various places for build
-
Martin Blix Grydeland authored
Format: <type> <parent-vxid> Add SLT_End records at the end of all vxid transactions. Add SLT_Link records whenever one initiates a new child transation. Format: <type-of-child> <child-vxid> This is used by VSL api to form structured log output.
-
Martin Blix Grydeland authored
-
Martin Blix Grydeland authored
process uptime counter.
-
Per Buer authored
Basically the backend can drop the reason (OK) when responding to a probe without Varnish declaring the backend sick. Reported by astrosti on https://www.varnish-cache.org/forum/topic/961
-
- 14 May, 2013 2 commits
-
-
Poul-Henning Kamp authored
first one created. I would have skipped #0 if it was me...
-
Poul-Henning Kamp authored
-
- 13 May, 2013 6 commits
-
-
Poul-Henning Kamp authored
bereq.* rather than the req.*, since that's what the backend saw.
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
only pass req to FetchHdr if we intend to send the body over.
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
Change args to http_FilterReq() to lower level.
-
Poul-Henning Kamp authored
-
- 12 May, 2013 2 commits
-
-
Poul-Henning Kamp authored
After the purge, you will end up in vcl_purge{} from where it will (at some future date) be possible to order a fresh fetch of the object. Doing purge from vcl_recv{} solves half of the object existence problems in vcl_lookup{}.
-
Poul-Henning Kamp authored
return (error(status_code [ , response_string ]));
-
- 08 May, 2013 1 commit
-
-
Per Buer authored
-
- 07 May, 2013 3 commits
-
-
Poul-Henning Kamp authored
-
Poul-Henning Kamp authored
next method we'll end up in.
-
Poul-Henning Kamp authored
-
- 06 May, 2013 3 commits
-
-
Per Buer authored
-
Per Buer authored
Fixes #1281
-
Poul-Henning Kamp authored
-