Commit Graph

65 Commits (f59e8bf5551f403e6b7ec0912097bce85ea21351)

Author SHA1 Message Date
Dmitry K e9d64ec29d Use `time.Equal` instead of direct comparison 2023-03-24 17:51:34 -05:00
Dmitry K 2f1bba09c4 Guard deadline readings by mutex 2023-03-24 17:51:34 -05:00
Dmitry K d829073b2f Improve deadline simulation 2023-03-24 17:51:34 -05:00
Dmitry K 48da6435a5 Add deadline simulation 2023-03-24 17:51:34 -05:00
Dmitry K 34e3013153 Remove commented out atomic calls 2023-03-24 17:51:34 -05:00
Dmitry K 009a377028 Use mutex to guard entire `SetBlockingMode` call 2023-03-24 17:51:34 -05:00
Dmitry K e05abb83ec Better error messages 2023-03-24 17:51:34 -05:00
Dmitry K 89475c4c91 use `atomic.Int32` instead of `int + atomic calls` 2023-03-24 17:51:34 -05:00
Dmitry K c3d62c8783 Small comment update 2 2023-03-24 17:51:34 -05:00
Dmitry K 1298a835bc Small comment update 2023-03-24 17:51:34 -05:00
Dmitry K b2b4fbcf57 Set socket to non-blocking mode in `Read`, `Flush` and `BufferReadUntilBlock` operations 2023-03-24 17:51:34 -05:00
Dmitry K a83faa67f5 Small improvements 2023-03-24 17:51:34 -05:00
Dmitry K 8b5e8d9d89 Fix Windows non-blocking I/O for CopyFrom
Created based on discussion here: https://github.com/jackc/pgx/pull/1525#pullrequestreview-1344511991

Fixes https://github.com/jackc/pgx/issues/1552
2023-03-24 17:51:34 -05:00
Nicola Murino 19039e6dd1 fix build on 32-bit Windows 2023-03-07 17:09:03 -06:00
Dmitry K 0dbb0a52ab Fix `realNonblockingRead`, set `realNonblockingRead` call error to `nonblockReadErr` 2023-03-04 09:25:36 -06:00
Dmitry K 087b8b2ba8 Try to make windows non-blocking I/O 2023-03-04 09:25:36 -06:00
Jack Christensen c09ddaf440 Add Windows non-blocking IO 2023-03-04 09:25:36 -06:00
Jack Christensen 6105ca5073 Fix TestInternalNonBlockingWriteWithDeadline(t
The test was relying on sending so big a message that the write blocked.
However, it appears that on Windows the TCP connections over localhost
have an very large or infinite sized buffer. Change the test to simply
set the deadline to the current time before triggering the write.
2023-02-25 17:02:55 -06:00
Jack Christensen 8f46c75e73 Fix: fake non-blocking read adaptive wait time
If the time reached the minimum time before the 5 tries were up it
would get stuck reading 1 byte at a time indefinitely.
2023-02-25 16:45:34 -06:00
Jack Christensen b707faea8f Fix flickering test TestBufferNonBlockingRead 2023-02-10 19:40:31 -06:00
Jack Christensen 9963c32d4f Only count when bytes actually read 2023-01-31 20:35:44 -06:00
Jack Christensen 6bc327b3ce Find fastest possible read time for fakeNonblockingReadWaitDuration
The first 5 fake non-blocking reads are limited to 1 byte. This should
ensure that there is a measurement of a read where bytes are already
waiting in Go or the OS's read buffer.
2023-01-31 20:25:57 -06:00
Jack Christensen f46d35610e Only set c.fakeNonblockingReadWaitDuration when it will be decreased 2023-01-31 20:25:17 -06:00
Jack Christensen cf78472ce5 Use unix build tag
With Go 1.19 available we can use a simpler build tag.
2023-01-31 20:10:34 -06:00
Jack Christensen 898891a6ee Fake non-blocking read adapts its max wait time
The reason for a high max wait time was to ensure that reads aren't
cancelled when there is data waiting for it in Go or the OS's receive
buffer. Unfortunately, there is no way to know ahead of time how long
this should take.

This new code uses 2x the fastest successful read time as the max read
time. This allows the code to adapt to whatever host it is running on.

https://github.com/jackc/pgx/issues/1481
2023-01-28 09:35:52 -06:00
Jack Christensen 7019ed1edf Fix tests for iobufpool optimization 2023-01-28 09:30:12 -06:00
Jack Christensen eee854fb06 iobufpool uses *[]byte instead of []byte to reduce allocations 2023-01-28 08:02:49 -06:00
Jack Christensen bc754291c1 Save memory on non blocking read path
Only create RawConn.Read callback once and have it use NetConn fields.
Avoids the closure and some allocations.

https://github.com/jackc/pgx/issues/1481
2023-01-27 20:53:30 -06:00
Jack Christensen 2c7d86a543 Only create RawConn.Write callback once
This saves an allocation on every call.

https://github.com/jackc/pgx/issues/1481
2023-01-27 20:34:21 -06:00
Jack Christensen 7941518809 BufferReadUntilBlock should release buf when no bytes read
This was causing allocations every time there was a non-blocking read
with nothing to read.

https://github.com/jackc/pgx/issues/1481
2023-01-27 18:03:38 -06:00
Alexey Palazhchenko f839d501a7 Apply `gofmt -s`
And add CI check for that.
2023-01-24 07:55:00 -06:00
Mark Chambers 516300aabf spelling: successfully, compatibility 2023-01-16 20:06:01 -06:00
Jack Christensen ba4bbf92af Fix query sanitizer
...when query text has contains Unicode replacement character.
uft8.RuneError actually is a valid character.
2022-11-14 18:32:26 -06:00
Jack Christensen d7c7ddc594 Fix Windows 386 atomic usage
https://github.com/jackc/pgx/issues/1307
2022-09-24 09:23:36 -05:00
Jack Christensen 906f709e0c Fix typo in Windows code
https://github.com/jackc/pgx/issues/1274
2022-08-11 20:59:37 -05:00
Jack Christensen 33b782a96d Potential fix for Windows
https://github.com/jackc/pgx/issues/1274
2022-08-11 20:55:50 -05:00
Jack Christensen a5b4f888c2 Fix flickering test on CI
Ensure the conn reads everything expected before closing.
2022-07-16 18:16:19 -05:00
Jack Christensen ca41a6a222 Update docs 2022-07-10 14:32:08 -05:00
Jack Christensen e7aa76ccf9 SendBatch now uses pipeline mode to prepare and describe statements
Previously, a batch with 10 unique parameterized statements executed
100 times would entail 11 network round trips. 1 for each prepare /
describe and 1 for executing them all. Now pipeline mode is used to
prepare / describe all statements in a single network round trip. So it
would only take 2 round trips.
2022-07-09 09:32:36 -05:00
Jack Christensen ed3e9f1dd4 Check for more specific error 2022-07-01 15:33:12 -05:00
Jack Christensen 9afd320b9e Fix flickering test in CI
While this test always worked on my machine, it flickered in CI. And to
be fair the test can't guarantee the condition it is testing. Work
around this by trying many times before admitting failure.
2022-06-25 16:05:20 -05:00
Jack Christensen 72b1dcff2f Add pgconn.CheckConn 2022-06-25 15:55:09 -05:00
Jack Christensen 811d855a35 Add non-blocking IO
This eliminates an edge case that can cause a deadlock and is a
prerequisite to cheaply testing connection liveness and to recoving a
connection after a timeout.

https://github.com/jackc/pgconn/issues/27

Squashed commit of the following:

commit 0d7b0dddea
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 25 13:15:05 2022 -0500

    Add test for non-blocking IO preventing deadlock

commit 79d68d23d3
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 18 18:23:24 2022 -0500

    Release CopyFrom buf when done

commit 95a43139c7
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 18 18:22:32 2022 -0500

    Avoid allocations with non-blocking write

commit 6b63ceee07
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 18 17:46:49 2022 -0500

    Simplify iobufpool usage

commit 60ecdda02e
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 18 11:51:59 2022 -0500

    Add true non-blocking IO

commit 7dd26a34a1
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 4 20:28:23 2022 -0500

    Fix block when reading more than buffered

commit afa702213f
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 4 20:10:23 2022 -0500

    More TLS support

commit 51655bf8f4
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 4 17:46:00 2022 -0500

    Steps toward TLS

commit 2b80beb1ed
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 4 13:06:29 2022 -0500

    Litle more TLS support

commit 765b2c6e7b
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 4 12:29:30 2022 -0500

    Add testing of TLS

commit 5b64432afb
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 4 09:48:19 2022 -0500

    Introduce testVariants in prep for TLS

commit ecebd7b103
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 4 09:32:14 2022 -0500

    Handle and test read of previously buffered data

commit 09c64d8cf3
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 4 09:04:48 2022 -0500

    Rename nbbconn to nbconn

commit 73398bc67a
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 4 08:59:53 2022 -0500

    Remove backup files

commit f1df39a29d
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 4 08:58:05 2022 -0500

    Initial passing tests

commit ea3cdab234
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat Jun 4 08:38:57 2022 -0500

    Fix connect timeout

commit ca22396789
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Thu Jun 2 19:32:55 2022 -0500

    wip

commit 2e7b46d5d7
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Mon May 30 08:32:43 2022 -0500

    Update comments

commit 7d04dc5caa
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat May 28 19:43:23 2022 -0500

    Fix broken test

commit bf1edc77d7
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat May 28 19:40:33 2022 -0500

    fixed putting wrong size bufs

commit 1f7a855b2e
Author: Jack Christensen <jack@jackchristensen.com>
Date:   Sat May 28 18:13:47 2022 -0500

    initial not quite working non-blocking conn
2022-06-25 13:15:31 -05:00
Jack Christensen e12ba1b6b9 Extract iobufpool 2022-05-28 10:59:54 -05:00
Jack Christensen 5714896b10 Restructure sending messages
Use an internal buffer in pgproto3.Frontend and pgproto3.Backend instead
of directly writing to the underlying net.Conn. This will allow tracing
messages as well as simplify pipeline mode.
2022-05-21 11:06:44 -05:00
Jack Christensen 1f4b34f932 Merge branch 'master' into v5-dev 2022-04-23 11:05:24 -05:00
Jack Christensen c323ab6662 Merge remote-tracking branch 'pgconn/master' into v5-dev 2022-04-23 10:48:44 -05:00
Jack Christensen f14fb3d692 Replace interface{} with any 2022-04-09 09:12:55 -05:00
Jack Christensen f27178ba85 Initial privatization of stmtcache
ConnConfig.BuildStatementCache is pending removal once connections
always have separate caches for prepared and described statements.
2022-03-12 08:35:31 -06:00
Jack Christensen 1cef9075d9 Simply typed nil and driver.Valuer handling
* Convert typed nils to untyped nils at beginning of encoding process.
* Restore v4 json/jsonb null behavior
* Add anynil internal package
2022-03-05 19:53:59 -06:00