Hi, Willy thanks for clarification about lfs_file, I've removed it from patch. As Christopher mentioned, when developing the patch I've noticed the comment that when "action_ptr" is set, "action" can be set to any meaningful value. However, I also had doubts about it. :)
Christopher I fixed the parsing issue (I hope), good catch and many thanks! sob., 21 lis 2020 o 10:43 Willy Tarreau <[email protected]> napisał(a): > On Sat, Nov 21, 2020 at 09:42:32AM +0100, Christopher Faulet wrote: > > ..Le 21/11/2020 à 06:15, Willy Tarreau a écrit : > > > I looked at the patch, it's OK regarding this and most of the other > > > stuff, except one thing, it mixes ACT_HTTP_* and PAT_MATCH_* and there > > > is some breakage there, because the HTTP action being set to > PAT_MATCH_STR > > > for example sets it to value 5 which is ACT_HTTP_SET_NICE, etc. So you > > > must only have ACT_* values in your rule->action field. > > > > > > Maciej, there are two possibilities here. Either you duplicate the > delete > > > action to have DEL_STR, DEL_BEG, DEL_SUB etc, and you just set the > right > > > action based on the matching method you found during parsing, or you > > > extend act_rule.arg.http to add a matching method for the delete > action. > > > I tend to consider that the former is cleaner and simpler, given that > > > we're really performing a few different actions there, which are all > > > derivatives of the delete action. > > > > Hi, > > > > rule->action value may be set to any value if an action function is > defined > > (rule->action_ptr). ACT_* are only used for action without action > function, > > to perform a switch case on it. That's documented this way : > > > > unsigned int action; /* ACT_* or any meaningful value if action_ptr is > defined */ > > OK, I wasn't aware of this, seeing only ACT_* there. Fine by me then. > > Willy >
0001-BUG-MINOR-http_htx-Fix-searching-headers-by-substrin.patch
Description: Binary data
0002-MINOR-http_act-Add-m-flag-for-del-header-name-matchi.patch
Description: Binary data
