Permalink
Commits on Feb 14, 2011
  1. OSX is dumb dumb dumb and reports even 64bit machines as i386

    Ezra Zygmuntowicz
    Ezra Zygmuntowicz committed Feb 14, 2011
Commits on Feb 12, 2011
  1. emphasized in README that Mac 64-bit users have to edit the "config" …

    agentzh committed Feb 12, 2011
    …file themselves when building with LuaJIT 2.0 (for now).
  2. minor tweaks of README.

    agentzh committed Feb 12, 2011
Commits on Feb 11, 2011
  1. now we always allocate r->request_body for subrequests when the metho…

    agentzh committed Feb 11, 2011
    …d option is specified for ngx.location.capture*. this prevents accidental inheritance of parent request's request body when client_body_buffer_size < client_max_body_size.
Commits on Feb 10, 2011
  1. fixed indentation.

    agentzh committed Feb 10, 2011
  2. fixed a typo in README.

    agentzh committed Feb 10, 2011
  3. added notes about variable number of subrequests for ngx.location.cap…

    agentzh committed Feb 10, 2011
    …ture_multi to README. thanks Marcus Clyne.
Commits on Feb 9, 2011
  1. optimized the buffer management process a bit for ngx.location.captur…

    agentzh committed Feb 9, 2011
    …e*: one the response body is a single buf chain, do not bother concatening it into a new buffer.
Commits on Feb 8, 2011
  1. now we no longer explicitly require OpenSSL crypto, we now use ngx_md…

    agentzh committed Feb 8, 2011
    …5.h exclusively, and no NDK_HASH pain any more.
Commits on Feb 6, 2011
Commits on Feb 5, 2011
  1. minor test adjust.

    agentzh committed Feb 5, 2011
  2. now we report an error at config time when rewrite_by_lua* are used w…

    agentzh committed Feb 5, 2011
    …ith nginx 0.8.42 ~ 0.8.53.
  3. minor README tweaks.

    agentzh committed Feb 5, 2011
Commits on Feb 4, 2011
  1. fixed regressions while building with nginx 0.7.x and also confirmed …

    agentzh committed Feb 4, 2011
    …that rewrite_by_lua* work with nginx 0.7.x.
  2. added notes regarding cases when ngx.location.capture and ngx.locatio…

    agentzh committed Feb 4, 2011
    …n.capture_multi fail to capture subrequests' outputs.
  3. now we recycle the chain links and associated bufs among subrequests …

    agentzh committed Feb 4, 2011
    …issued by one or more ngx.location.capture_multi() calls within a single (parent) request.
  4. added some tests for using ngx.location.capture_multi() at rewrite, a…

    agentzh committed Feb 4, 2011
    …ccess, and content phases at the same time and these tests are all passing :)
  5. we now cancel the ctx->body field and use the parent request's ctx->b…

    agentzh committed Feb 4, 2011
    …odies instead, thus saving some bytes per request.
  6. we now post the parent request ourselves in the post_subrequest callb…

    agentzh committed Feb 4, 2011
    …ack in some extreme condition that the current (sub)request is not active AND it has some weird postponed requests associated with it. this work-around makes those crazily complicated test cases for ngx.location.capture_multi() pass now.
  7. minor optimizations: we only pass NULL chains down the output filter …

    agentzh committed Feb 4, 2011
    …chain when we really need (that is, the current request is active *AND* there is postponed outputs in r->postponed).
Commits on Feb 3, 2011
  1. minor optimizations in ngx.location.capture_multi() and also fixed a …

    agentzh committed Feb 3, 2011
    …bug in its request options handling (we did not initialize the options properly for *every* request).
  2. the post subrequest callback may be called multiple times (due to fin…

    agentzh committed Feb 3, 2011
    …alizing in-active requests) and those useless left-over calls may cause segfaults while accessing parent requests' ctx struct. now we add checks to post_subrequest to prevent its main logic from running twice or more.