# HG changeset patch
# User Piotr Sikora
# Date 1467064805 25200
# Mon Jun 27 15:00:05 2016 -0700
# Node ID 3ba6126630ceec0767e70b88be1a004d82e2c27e
# Parent d452cb27639f714bc43b7f3ede417e3e7f8efdd6
Configure: fix build with -Werror=old-style-definition.
Signed-off-by: Piotr Sikora
diff -
# HG changeset patch
# User Piotr Sikora
# Date 1467064806 25200
# Mon Jun 27 15:00:06 2016 -0700
# Node ID 388cf446fe903cfdf0d13740ab3489cc14a3f377
# Parent d452cb27639f714bc43b7f3ede417e3e7f8efdd6
Configure: stop polluting NGX_ namespace.
While there, fix the only test that used alternati
# HG changeset patch
# User Piotr Sikora
# Date 1467064807 25200
# Mon Jun 27 15:00:07 2016 -0700
# Node ID 8666abaeb43294edd5b93852bad2a42a27dd651b
# Parent d452cb27639f714bc43b7f3ede417e3e7f8efdd6
Core: remove unused NGX_INVALID_ARRAY_INDEX.
Unused since ae4744c28620 (0.7.51).
Signed-off
Hey Shuxin,
> I took some time to read your code, I believe I fully understand your
> code at this moment.
> My implementation is slightly differently from yours, and we have different
> use-cases.
Thanks!
> Here is my $0.02 value of comment to your code (cosmetic wise):
> - the v
It's "trailer support" for the web serving part of NGINX, just not proxying.
- the "get" in "ngx_http_chunked_get_trailers" is sort of misleading.
The "get" here actually
means "generate". I thought you are looking for a existing buffer.
That function name got changed a few time
# HG changeset patch
# User Piotr Sikora
# Date 1467064809 25200
# Mon Jun 27 15:00:09 2016 -0700
# Node ID 1059f59c44039020843ced5bca39f165c322bacb
# Parent d452cb27639f714bc43b7f3ede417e3e7f8efdd6
Core: remove unused NGX_SIG_ATOMIC_T_SIZE.
Unused since 9c2f3ed7a247 (0.3.3).
Signed-off-by
# HG changeset patch
# User Piotr Sikora
# Date 1467064810 25200
# Mon Jun 27 15:00:10 2016 -0700
# Node ID a2892701d91e6aba62719b54792e1e14eeea11a0
# Parent 1059f59c44039020843ced5bca39f165c322bacb
Core: remove NGX_TIME_T_SIZE.
Replace ./configure-time test with equivalent compile-time tes
Hey Shuxin,
> How do you manage to tackle this scenario without changing your existing
> code:
>
> backend server responses with trailer "lol = 123", and you define the
> same trailer,
> "lol = $haha" in *.conf. You have two options as to how to combine them:
> either override the
> one from o
# HG changeset patch
# User Piotr Sikora
# Date 1467064811 25200
# Mon Jun 27 15:00:11 2016 -0700
# Node ID 11a53446bd90b8927cb93847c98672a8f056d182
# Parent d452cb27639f714bc43b7f3ede417e3e7f8efdd6
Core: reorder #ifdef in .h file to match order in .c file.
Signed-off-by: Piotr Sikora
dif
# HG changeset patch
# User Piotr Sikora
# Date 1467064812 25200
# Mon Jun 27 15:00:12 2016 -0700
# Node ID e4375e338d9acf927ffdc4cd18c704bb7ea1fd4e
# Parent d452cb27639f714bc43b7f3ede417e3e7f8efdd6
Configure: don't test alternatives if preferred feature is found.
Signed-off-by: Piotr Sikor
Hello,
I stuck in rather strange situation: if I run nginx under debugger, and
request zero-length file it gives me correct reply on first connection. But
on second connection it reply it output part of debug log (what supposed to
be written to debug log, but wasn't).
At the same time if nginx run
___
nginx-devel mailing list
nginx-devel@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx-devel
Hello!
On Mon, Jun 27, 2016 at 12:30:19PM +0300, Andrey Kulikov wrote:
> I stuck in rather strange situation: if I run nginx under debugger, and
> request zero-length file it gives me correct reply on first connection. But
> on second connection it reply it output part of debug log (what supposed
Hello!
On Sun, Jun 26, 2016 at 04:13:15PM -0700, Piotr Sikora wrote:
> # HG changeset patch
> # User Piotr Sikora
> # Date 1466735480 25200
> # Thu Jun 23 19:31:20 2016 -0700
> # Node ID a96187a9806536ab2a8bf3fa7f7b659cf5d3ff13
> # Parent 6f69e3c0f780e29bca752fc1f938f4a459a1ec59
> HTTP: ad
Hey Maxim,
> What's the goal? Any real-world use for this?
For us? gRPC, which uses HTTP/2 and/or HTTP/1.1 with trailers as a
transport protocol.
> As far as I understand the patch, this will cause chunked encoding
> to be used for all responses to a client which supports trailers.
> This certa
details: http://hg.nginx.org/nginx/rev/d452cb27639f
branches:
changeset: 6601:d452cb27639f
user: Roman Arutyunyan
date: Mon Jun 27 18:42:29 2016 +0300
description:
Style.
diffstat:
src/stream/ngx_stream_limit_conn_module.c | 20 ++--
1 files changed, 10 insertions
Hi Maxim,
Thanks for a fast reply!
Indeed, I didn't mean to reload configuration. After instructing gdb not to
forward SIGHUP to the process being debugged, everything start to work like
a charm.
Thanks for help!
--
WBR,
Andrey
On 27 June 2016 at 16:34, Maxim Dounin wrote:
> Hello!
>
> On Mon
Hi, Piotr:
I'm wondering why not just change the ngx_http_chunked_filter_module.c?
or add a module inserted right after the chunked-filter-module?
Shuxin
On 06/27/2016 07:37 AM, Piotr Sikora wrote:
Hey Maxim,
What's the goal? Any real-world use for this?
On 06/27/2016 06:45 AM, Maxim
Hey Shuxin,
>I'm wondering why not just change the ngx_http_chunked_filter_module.c?
> or add a module inserted right after the chunked-filter-module?
Hmm...? I'm confused by your comment.
This patch modifies 3 files:
- parsing of "TE" header is done in ngx_http_request.c,
- HTTP/1.1 trailer
Hello!
On Mon, Jun 27, 2016 at 07:37:31AM -0700, Piotr Sikora wrote:
> > What's the goal? Any real-world use for this?
>
> For us? gRPC, which uses HTTP/2 and/or HTTP/1.1 with trailers as a
> transport protocol.
Last time I checked gRPC wasn't able to talk HTTP/1.1, and
therefore wasn't able
Hi, Piotr and Maxim:
> to Maxim's question about "does trailer has real-world use" first.
At least we do, some info is available only after the entire body is
sent. Trailer is a viable and convenient
solution for that matter, albeit it is not the only solution for it. I
cannot "leak" more inf
Hey Maxim,
> Last time I checked gRPC wasn't able to talk HTTP/1.1, and
> therefore wasn't able to work with nginx at all. So no real-world
> use for now, right?
Official clients & servers support only HTTP/2, but gRPC protocol can
be layered on top of HTTP/1.1 with trailers, so it's up to the
i
Hey Shuxin,
> As far as I can understand, your change is just to add trailer headers
> (not including the part that paring incoming
> trailer header from upstream, or merge the incoming trailer and generated
> trailer). If that is correct, you just need
> to add "trailer: hdr1,hdr2... hdrn" to
Hi, Piotr:
I took some time to read your code, I believe I fully understand
your code at this moment.
My implementation is slightly differently from yours, and we have
different use-cases.
Here is my $0.02 value of comment to your code (cosmetic wise):
- the verb "support" in "
# HG changeset patch
# User Piotr Sikora
# Date 1467064802 25200
# Mon Jun 27 15:00:02 2016 -0700
# Node ID f5ece83dbcc4c46bcda51511e4fbf8fb90aa9432
# Parent d452cb27639f714bc43b7f3ede417e3e7f8efdd6
Configure: fix build with -Werror=unused-value.
Signed-off-by: Piotr Sikora
diff -r d452cb
# HG changeset patch
# User Piotr Sikora
# Date 1467064803 25200
# Mon Jun 27 15:00:03 2016 -0700
# Node ID ad603262a99df74951b05ce30c81ebc381694162
# Parent d452cb27639f714bc43b7f3ede417e3e7f8efdd6
Configure: fix build with -Werror=unused-but-set-variable.
Signed-off-by: Piotr Sikora
dif
# HG changeset patch
# User Piotr Sikora
# Date 1467064801 25200
# Mon Jun 27 15:00:01 2016 -0700
# Node ID d183012fa3d130b8de922d1c8eafd77ae94f8a51
# Parent d452cb27639f714bc43b7f3ede417e3e7f8efdd6
Configure: style.
Signed-off-by: Piotr Sikora
diff -r d452cb27639f -r d183012fa3d1 auto/un
# HG changeset patch
# User Piotr Sikora
# Date 1467064804 25200
# Mon Jun 27 15:00:04 2016 -0700
# Node ID 2c14d7fb1d96c3a1bfb7e22383bef895287bf7d7
# Parent d452cb27639f714bc43b7f3ede417e3e7f8efdd6
Configure: fix build with -Werror=nonnull.
Signed-off-by: Piotr Sikora
diff -r d452cb27639
# HG changeset patch
# User Piotr Sikora
# Date 1467064808 25200
# Mon Jun 27 15:00:08 2016 -0700
# Node ID 890dc5e6a88c49a9d724cc7da1d0eb20fcc70768
# Parent d452cb27639f714bc43b7f3ede417e3e7f8efdd6
Core: remove unused ngx_tm_zone.
Unused since 9c2f3ed7a247 (0.3.3).
Signed-off-by: Piotr Si
29 matches
Mail list logo