[nginx-tests] Tests: avoid changing non-localized $TODO.

2024-05-30 Thread Maxim Dounin
details: http://freenginx.org/hg/nginx-tests/rev/0e2b14c75232 branches: changeset: 1981:0e2b14c75232 user: Maxim Dounin date: Fri May 31 06:22:48 2024 +0300 description: Tests: avoid changing non-localized $TODO. This ensures that there will be no unrelated effects if the variable i

[nginx-tests] Tests: explicit Valgrind support.

2024-05-30 Thread Maxim Dounin
details: http://freenginx.org/hg/nginx-tests/rev/fb25cbe9d4ec branches: changeset: 1982:fb25cbe9d4ec user: Maxim Dounin date: Fri May 31 06:23:00 2024 +0300 description: Tests: explicit Valgrind support. Valgrind logging is done to a separate file, as it is not able to follow stderr

[nginx-tests] Tests: logging of certificate subjects instead of ...

2024-05-30 Thread Maxim Dounin
details: http://freenginx.org/hg/nginx-tests/rev/d732a96e05df branches: changeset: 1980:d732a96e05df user: Maxim Dounin date: Sat May 25 03:44:55 2024 +0300 description: Tests: logging of certificate subjects instead of issuers. diffstat: lib/Test/Nginx.pm| 2 +- lib/Test

[nginx-site] Added links to Mercurial repositories on download p...

2024-05-30 Thread Maxim Dounin
details: http://freenginx.org/hg/nginx-site/rev/329fb2b5e16b branches: changeset: 3085:329fb2b5e16b user: Maxim Dounin date: Tue May 28 00:23:26 2024 +0300 description: Added links to Mercurial repositories on download page. Non-linked repo URLs, as well as "read-only" in the descri

[nginx-site] Added link to Github mirror.

2024-05-30 Thread Maxim Dounin
details: http://freenginx.org/hg/nginx-site/rev/40d3fc1cccdb branches: changeset: 3086:40d3fc1cccdb user: Maxim Dounin date: Tue May 28 00:23:28 2024 +0300 description: Added link to Github mirror. diffstat: xml/en/download.xml | 7 ++- xml/ru/download.xml | 7 ++- 2 fil

[nginx] Removed incorrect NGX_SUPPRESS_WARN usage.

2024-05-30 Thread Maxim Dounin
details: http://freenginx.org/hg/nginx/rev/2fce021888f9 branches: changeset: 9279:2fce021888f9 user: Maxim Dounin date: Fri May 31 04:38:40 2024 +0300 description: Removed incorrect NGX_SUPPRESS_WARN usage. The len variable in ngx_http_variable_unknown_header() needs to be initializ

[nginx] Fixed Valgrind complaints about uninitialized values.

2024-05-30 Thread Maxim Dounin
details: http://freenginx.org/hg/nginx/rev/f53146df9a47 branches: changeset: 9278:f53146df9a47 user: Maxim Dounin date: Fri May 31 04:38:09 2024 +0300 description: Fixed Valgrind complaints about uninitialized values. In ngx_http_source_charset(), name->data was left uninitialized,

[PATCH 7 of 7] HTTP/3: protection from recursion during connection reuse

2024-05-30 Thread Maxim Dounin
# HG changeset patch # User Maxim Dounin # Date 1717116498 -10800 # Fri May 31 03:48:18 2024 +0300 # Node ID 5710b269079c37a0bc99fcfc0e156f971fdfe75a # Parent 8eecd832d2711dca1f7d4eff96369cdc3be514f4 HTTP/3: protection from recursion during connection reuse. When draining a connection assoc

[PATCH 5 of 7] QUIC: ngx_quic_buffer_t use-after-free protection

2024-05-30 Thread Maxim Dounin
# HG changeset patch # User Roman Arutyunyan # Date 1716902361 -14400 # Tue May 28 17:19:21 2024 +0400 # Node ID 9e69f2b7026deeed2824422ef64d11e234c05bb1 # Parent 8822af43e7e919b8c903174886c8ffecdb25f017 QUIC: ngx_quic_buffer_t use-after-free protection. Previously the last chain field of n

[PATCH 4 of 7] QUIC: ignore CRYPTO frames after handshake completion

2024-05-30 Thread Maxim Dounin
# HG changeset patch # User Roman Arutyunyan # Date 1716902348 -14400 # Tue May 28 17:19:08 2024 +0400 # Node ID 8822af43e7e919b8c903174886c8ffecdb25f017 # Parent 352360116e2c2fef891a91284083fb1d6c36c72d QUIC: ignore CRYPTO frames after handshake completion. Sending handshake-level CRYPTO f

[PATCH 6 of 7] HTTP/3: fixed handling of zero-length literal field line

2024-05-30 Thread Maxim Dounin
# HG changeset patch # User Sergey Kandaurov # Date 1716902445 -14400 # Tue May 28 17:20:45 2024 +0400 # Node ID 8eecd832d2711dca1f7d4eff96369cdc3be514f4 # Parent 9e69f2b7026deeed2824422ef64d11e234c05bb1 HTTP/3: fixed handling of zero-length literal field line. Previously, st->value was pas

[PATCH 3 of 7] HTTP/3: fixed dynamic table overflow

2024-05-30 Thread Maxim Dounin
# HG changeset patch # User Roman Arutyunyan # Date 1716902330 -14400 # Tue May 28 17:18:50 2024 +0400 # Node ID 352360116e2c2fef891a91284083fb1d6c36c72d # Parent f9bd4667574b1e6b36a04eb0769c510235df5488 HTTP/3: fixed dynamic table overflow. While inserting a new entry into the dynamic tabl

[PATCH 1 of 7] QUIC: fixed close timer processing with early data

2024-05-30 Thread Maxim Dounin
# HG changeset patch # User Vladimir Khomutov # Date 1712731090 -10800 # Wed Apr 10 09:38:10 2024 +0300 # Node ID 9e9988d46171cfe26fe0f9d24a51eb38926adbb8 # Parent 4a0cd107c0f1d56cd455df39c37eb0fffc085760 QUIC: fixed close timer processing with early data. The ngx_quic_run() function uses q

[PATCH 2 of 7] QUIC: client transport parameter data length checking

2024-05-30 Thread Maxim Dounin
# HG changeset patch # User Sergey Kandaurov # Date 1716902239 -14400 # Tue May 28 17:17:19 2024 +0400 # Node ID f9bd4667574b1e6b36a04eb0769c510235df5488 # Parent 9e9988d46171cfe26fe0f9d24a51eb38926adbb8 QUIC: client transport parameter data length checking. diff --git a/src/event/quic/ngx_

[PATCH 0 of 7] QUIC / HTTP/3 fixes merge

2024-05-30 Thread Maxim Dounin
Hello! The following patch series merges QUIC / HTTP/3 fixes from F5. Additionally, there is a (different) fix for the HTTP/3 connection reuse issue. Comments are welcome. -- Maxim Dounin

Re: proxy_cache_lock for content revalidation

2024-05-30 Thread Maxim Dounin
Hello! On Thu, May 30, 2024 at 01:05:08PM -0700, Maksim Yevmenkin wrote: > it seems that the proxy_cache_lock directive operates only for cache > misses (new content). while this behavior is documented, i am curious > about the reasoning behind it. there are scenarios where > proxy_cache_lock cou

proxy_cache_lock for content revalidation

2024-05-30 Thread Maksim Yevmenkin
hello! it seems that the proxy_cache_lock directive operates only for cache misses (new content). while this behavior is documented, i am curious about the reasoning behind it. there are scenarios where proxy_cache_lock could be very beneficial for content revalidation. what are the community's th