# HG changeset patch
# User Thibault Charbonnier
# Date 1592445047 25200
# Wed Jun 17 18:50:47 2020 -0700
# Node ID 3b1e0c7867c40fc82455b5c5dbf1d2e1521c1c8b
# Parent 8cf31489b479b689b7ff4a9601ce24c914d0394c
Fixed use-of-uninitialized-value inside ngx_epoll_test_rdhup.
An issue detected by
On 4/27/20 4:26 PM, Thibault Charbonnier wrote:
> Great! Thanks for the suggestion. Below is a revised approach for the
> patch (also attached to this email) which passes all of the test cases
> listed in my previous test file at the start of this thread:
Hello,
Are there plans for a
tached to this email) which passes all of the test cases
listed in my previous test file at the start of this thread:
# HG changeset patch
# User Thibault Charbonnier
# Date 1582764433 28800
# Wed Feb 26 16:47:13 2020 -0800
# Node ID 8d781bac6c4feebb2d1ea3f4e6df76d71f7
On 3/3/20 6:28 AM, Maxim Dounin wrote:
> Checking for the oldpid file does not look like a reliable
> approach to me.
Hi Maxim,
For this patch to cover TEST 3 in my previously attached test suite
(cancel binary upgrade via SIGQUIT on the new binary), we need a way for
a new binary to know whet
Hello Maxim,
Any thoughts on the latest version of this patch?
Cheers,
On 2/27/20 4:23 PM, Thibault Charbonnier wrote:
> On 2/27/20 7:24 AM, Maxim Dounin wrote:
>> Have you checked what happens during binary upgrade with your
>> patch?
>
> Good call, I gave it a thoug
binary upgrade edge-cases
by relying on the existence of the nginx.oldpid file.
Also attached to this email is a file I used as a test suite covering
the behavior of this patch with SIGQUIT, SIGTERM, and binary upgrade
scenarios.
# HG changeset patch
# User Thibault Charbonnier
# Date 1582764433
# HG changeset patch
# User Thibault Charbonnier
# Date 1582764433 28800
# Wed Feb 26 16:47:13 2020 -0800
# Node ID 55ea1a9197a6f28d4da00909e5ea8585f6a08239
# Parent 4f18393a1d51bce6103ea2f1b2587900f349ba3d
Ensured SIGQUIT deletes listening UNIX socket files.
Prior to this patch, the
# HG changeset patch
# User Thibault Charbonnier
# Date 1567537546 25200
# Tue Sep 03 12:05:46 2019 -0700
# Node ID 68ba3d36bff4213e3fedc538021e8cbece85e508
# Parent 52b5ee64fe11ec267a0767cbb9874c8cae652299
HTTP: added the preserve_method option to the error_page directive.
As of today
# HG changeset patch
# User Thibault Charbonnier
# Date 1567193727 25200
# Fri Aug 30 12:35:27 2019 -0700
# Node ID 40abd582aafbd22c2435afed5fd3311333ca99bd
# Parent a31ec2b79f9f208fecbc92020d12edb2a5207480
Upstream keepalive: keepalive_pool_key directive.
This directive gives control to
Hi,
Thank you Ruslan for your answer.
On 9/14/17 6:31 AM, Ruslan Ermilov wrote:
I personally don't like the idea of parsing /etc/hosts by nginx.
The approach choosen looks like a hack, as obviously editing the
/etc/hosts file between configuration reloads won't take any effect,
which is controv
from being merged, I would be glad to improve
it (for now, the tests have been written with Test::Nginx only).
Best,
Thibault
# HG changeset patch
# User Thibault Charbonnier
# Date 1488252201 28800
# Mon Feb 27 19:23:21 2017 -0800
# Node ID 558041ef1d70689ccc4c12f2487c3f75e6bbbcc
Hi,
Just curious what is the community's stance on this? If the feature is
desired but the patch isn't good, I am willing to receive feedback and
improve it. If not desired, may I be so curious as to ask why?
I hope the ping is considered ok.
Best,
Thibault
__
On 2/27/17 7:45 PM, Thibault Charbonnier wrote:
+if (r->hostsfile.len > 0
+&& ngx_resolver_parse_hostsfile(cf, r, r->hostsfile)
My bad, the third argument here can probably be removed. I moved the
hosts file path to the resolver struct at some point in order
the hosts file
is used for a particular resolver or not
Please let me know if this is of any interest.
--
Thibault
# HG changeset patch
# User Thibault Charbonnier
# Date 1488252201 28800
# Mon Feb 27 19:23:21 2017 -0800
# Branch resolve-hostsfile
# Node ID
On 2/1/17 8:58 AM, Maxim Dounin wrote:
I'm highly sceptical about introducing variables support
everywhere just to save some configuration complexity in very
special cases. In most cases there better ways to organize things
without using variables, as initial nginx configuration approach
suggest
Hello
> On Feb 1, 2017, at 4:54 AM, Maxim Dounin wrote:
>
> until there is a way to introduce variables
> support with less effort.
What do you suggest to reduce the patch size? Should it take care of those
other timeout directives in other modules?
One of my other hunch was to add the com
Hi!
On 1/31/17 8:00 AM, Maxim Dounin wrote:
You may want to provide more details on what problem you are
trying to solve.
The use-case was originally to use a longer proxy_read_timeout on
certain upstream locations without duplicating too much configuration
(our location block). I then thoug
patch
# User Thibault Charbonnier
# Date 1485823618 28800
# Mon Jan 30 16:46:58 2017 -0800
# Branch proxy/variables-in-timeouts
# Node ID 9f27c1d40590566ae5ac55e04af473d5f8e97d87
# Parent 640f035293959b2d4b0ba5939d954bc517f57f77
Proxy: make timeout directives accept variables
We try to stay as
Hello!
Quite minor, but as far as I can tell, this field is not being used
anywhere.
# HG changeset patch
# User Thibault Charbonnier
# Date 1485309159 28800
# Tue Jan 24 17:52:39 2017 -0800
# Branch remove-unused-upstream-struct-field
# Node ID bb5d6b4762998d1ca8f9ad487fa0e9bf6e0094e9
On 12/16/16 7:07 AM, Maxim Dounin wrote:
Hello!
Hi,
Builds without IPv6 support is something we are phasing out.
Starting with nginx 1.11.5 IPv6 support is automatically
configured as long as supported by OS:
Changes with nginx 1.11.5, 11 Oct 2016:
*) Change: the --with-ipv6 configure op
Hello,
Please let me know how you feel about this relatively simple patch which
aims at making the "resolver" directive more robust with regards to the
"ipv6=" option when Nginx is built without IPv6 support.
Thanks!
# HG changeset patch
# User Thibault Charbonnier
# D
21 matches
Mail list logo
