[issue41815] SQLite: segfault if backup called on closed database
New submission from Peter McCormick :
Attempting to backup a closed database will trigger segfault:
```
target = sqlite.connect(':memory:')
source = sqlite.connect(":memory:")
source.close()
source.backup(target)
```
--
files: fix.patch
keywords: patch
messages: 377176
nosy: pdmccormick
priority: normal
severity: normal
status: open
title: SQLite: segfault if backup called on closed database
type: crash
versions: Python 3.10, Python 3.7, Python 3.8, Python 3.9
Added file: https://bugs.python.org/file49464/fix.patch
___
Python tracker
<https://bugs.python.org/issue41815>
___
___
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue41815] SQLite: segfault if backup called on closed database
Peter McCormick added the comment:
Updated sample script, thanks to @aeros for catching the omission:
```
import sqlite3
target = sqlite3.connect(':memory:')
source = sqlite3.connect(':memory:')
source.close()
source.backup(target)
```
--
___
Python tracker
<https://bugs.python.org/issue41815>
___
___
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue28909] Adding LTTng-UST tracing support
Peter McCormick added the comment: I would like to propose the following patch[0] which generalizes the `PyDTrace_*` definitions to a more generic `PyProbe` structure, in anticipation of including LTTng support side-by-side with the existing DTrace support. A couple of argument types in `Include/pydtrace.h` were also changed to match those in `Include/pydtrace.d`. I can split that out to a separate, pre-PyProbe patch if that would be preferred. Some of the `*_ENABLED()` calls were reorganized such that they will actually be invoked twice (specifically in `Python/ceval.c` were the enabled check was used prior to calling helper functions, which then eventually invoked the DTrace macros.) If this won't work, we could add a sequence of guarded-by-ifdef-instead-of-enabled calls to ensure that there will only ever be one instance of any given `PyDTrace_*_ENABLED()` call. I've tested compilation on macOS with and without DTrace configured. Feedback and comments would be most welcome! Thanks, Peter [0] https://github.com/pdmccormick/cpython/commit/1e399f5ec381276b52e6a4f5a755fc0f23bd6d97 [1] https://github.com/pdmccormick/cpython/tree/bpo-28909-pyprobe-refactor -- versions: +Python 3.8 -Python 3.7 Added file: https://bugs.python.org/file47715/bpo-28909-pyprobe-refactor-20180727.patch ___ Python tracker <https://bugs.python.org/issue28909> ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue28909] Adding LTTng-UST tracing support
Peter McCormick added the comment:
Fork at
<https://github.com/pdmccormick/cpython/tree/bpo-28909-adding-lttng-ust> with
Francis' original patch, plus a revert for the DTrace provider name change.
Here are instructions I used to try this out:
## LTTng installation
Full installation instructions for LTTng are available at
<https://lttng.org/download/>. On Ubuntu 19.04 you can currently install LTTng
v2.7 from the `universe` repository if it is enabled:
```shell
$ apt-get install lttng-tools lttng-modules-dkms liblttng-ust-dev babeltrace
```
## Building
Clone/fetch/checkout from
<https://github.com/pdmccormick/cpython/tree/bpo-28909-adding-lttng-ust> and
run:
```shell
$ ./configure --with-lttngust
$ make
```
## Tracing a session
LTTng operates around the concept of tracing selected userspace & kernelspace
events of interest within the context of a session, so let's start one and run
the instrumented interpreter to see what happens:
```shell
$ lttng create 'pysession'
$ lttng enable-event --userspace 'cpython:*'
$ lttng start
$ ./python -m this
$ lttng stop
$ lttng view | wc -l
48257
$ lttng view
[23:50:04.493873655] (+?.?) keflavik cpython:gc__start: { cpu_id = 1 },
{ generation = 0 }
[23:50:04.493910472] (+0.36817) keflavik cpython:gc__done: { cpu_id = 1 },
{ collected = 0 }
[23:50:04.494281326] (+0.000370854) keflavik cpython:gc__start: { cpu_id = 1 },
{ generation = 0 }
[23:50:04.494305307] (+0.23981) keflavik cpython:gc__done: { cpu_id = 1 },
{ collected = 0 }
[23:50:04.495031049] (+0.000725742) keflavik cpython:gc__start: { cpu_id = 1 },
{ generation = 0 }
[23:50:04.495074272] (+0.43223) keflavik cpython:gc__done: { cpu_id = 1 },
{ collected = 0 }
[23:50:04.495403759] (+0.000329487) keflavik cpython:function__entry: { cpu_id
= 1 }, { co_filename = "", co_name = "",
line_no = 8 }
[23:50:04.495405056] (+0.01297) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "", line_no =
8 }
[23:50:04.495406486] (+0.01430) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "", line_no =
25 }
[23:50:04.495407149] (+0.00663) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "", line_no =
27 }
[23:50:04.495408921] (+0.01772) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "", line_no =
35 }
[23:50:04.495409618] (+0.00697) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "", line_no =
42 }
[23:50:04.495410775] (+0.01157) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "", line_no =
44 }
[23:50:04.495412006] (+0.01231) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "", line_no =
47 }
[23:50:04.495421990] (+0.09984) keflavik cpython:function__entry: { cpu_id
= 1 }, { co_filename = "", co_name =
"_DeadlockError", line_no = 47 }
[23:50:04.495422496] (+0.00506) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "_DeadlockError",
line_no = 47 }
[23:50:04.495423317] (+0.00821) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "_DeadlockError",
line_no = 48 }
[23:50:04.495424559] (+0.01242) keflavik cpython:function__return: { cpu_id
= 1 }, { co_filename = "", co_name =
"_DeadlockError", line_no = 48 }
[23:50:04.495468551] (+0.43992) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "", line_no =
51 }
[23:50:04.495470096] (+0.01545) keflavik cpython:function__entry: { cpu_id
= 1 }, { co_filename = "", co_name =
"_ModuleLock", line_no = 51 }
[23:50:04.495470570] (+0.00474) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "_ModuleLock", line_no
= 51 }
[23:50:04.495471249] (+0.00679) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "_ModuleLock", line_no
= 55 }
[23:50:04.495471746] (+0.00497) keflavik cpython:line: { cpu_id = 1 }, {
co_filename = "", co_name = "_ModuleLock", line_no
= 57 }
...
```
The raw trace files will be written under
`~/lttng-traces/pysession-$TIMESTAMP`. To clean them up afterwards, you can run:
```shell
$ lttng destroy
```
## GitHub
I would be happy to re-post this to GitHub issues if so desired.
--
hgrepos: +367
nosy: +pdmccormick
___
Python tracker
<http://bugs.python.org/issue28909>
___
___
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue28909] Adding LTTng-UST tracing support
Peter McCormick added the comment:
A few suggestions:
* Disallow `--with-lttngust` on anything other than Linux (on macOS
`configure` dies due to differences in acceptable `mktemp` arguments if you
even attempt it)
* Rename `cpython_inst.h` to `pytrace.h` and rename `pylttngust_probes.h` to
`pytrace_lttngust_probes.h`
- While we're at it, rename `pydtrace.h` to `pytrace_dtrace.h`, and
other mentions of `PyDTrace_*` to `PyTrace_DTrace_*`?
* Define a `WITH_TRACE` or `Py_TRACE` or similar preprocessor symbol
* Using that symbol, in `Python/ceval.c` ifdef-guard the static function
prototypes and function calls
- Otherwise when BOTH DTrace and LTTng-UST are disabled, Clang on macOS
gives `warning: code will never be executed` warnings on the various arms of
the `if (PyTraceEnabled(...))` statements, and GCC on Linux warn about unused
variables `lineno`, `funcname` and `filename` in
`pytrace_function_{entry,return}`, since the actual use of those variables as
arguments is preprocessed out of existance
If everyone was in agreement, would it make sense to sequence this as first the
generalization-renames to the existing DTrace/SystemTap code, and then recast
the LTTng addition patch on top of those? I'd be happy to do this.
So, `PyTrace` or `PyTracing`?
--
___
Python tracker
<http://bugs.python.org/issue28909>
___
___
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue28909] Adding LTTng-UST tracing support
Peter McCormick added the comment:
Hi Łukasz, thank you for the feedback!
> "PyTrace" is already a name in use for a different purpose. I understand the
> itch to make the name more "right" but I am in general not a fan of renaming
> "PyDTrace" to anything else now. It was a placeholder from day one (SystemTap
> uses it, too, after all). So, looking closer at the patch now I'd prefer us
> to keep all existing names and add LTTng as another alternative engine here.
> That will make the patch much shorter.
What about `PyProbe`? Given the multitude of tools and techniques in this
space, wouldn't it be worthwhile to clarify things before adding this? I think
conflating `dtrace` and `lttng` would only lead to more confusion for users as
they really are distinct technologies.
Apart from the `--with{out)-dtrace` configure options, are these terms exposed
to users anywhere else? I agree that those options shouldn't be changed now.
--
___
Python tracker
<http://bugs.python.org/issue28909>
___
___
Python-bugs-list mailing list
Unsubscribe:
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue10933] Tracing disabled when a recursion error is triggered (even if properly handled)
Changes by Peter McCormick : -- nosy: +pdmccormick ___ Python tracker <http://bugs.python.org/issue10933> ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue10933] Tracing disabled when a recursion error is triggered (even if properly handled)
Peter McCormick added the comment: Removed unrelated doc changes. -- Added file: http://bugs.python.org/file38909/issue_10933-2.patch ___ Python tracker <http://bugs.python.org/issue10933> ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue20174] Derby #5: Convert 50 sites to Argument Clinic across 3 files
Peter McCormick added the comment: I am working on revising the Argument Clinic definitions for socketmodule.c. -- nosy: +pdmccormick ___ Python tracker <http://bugs.python.org/issue20174> ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue23935] Clean up Clinic's type expressions of buffers
Changes by Peter McCormick : -- nosy: +pdmccormick ___ Python tracker <http://bugs.python.org/issue23935> ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue23967] Make inspect.signature expression evaluation more powerful
Peter McCormick added the comment: This definitely works for the _socket.listen use case! In terms of generating such a signature using Argument Clinic, currently this is required: backlog: int(py_default="builtins.min(SOMAXCONN, 128)", c_default="Py_MIN(SOMAXCONN, 128)") = 000 The attached patch lets Tools/clinic/clinic.py make an exception when both C and Python defaults are specified, simplifying the above to: backlog: int(py_default="builtins.min(SOMAXCONN, 128)", c_default="Py_MIN(SOMAXCONN, 128)") -- keywords: +patch Added file: http://bugs.python.org/file39066/pdm-argument_clinic-mixed_py_and_c_defaults-v1.patch ___ Python tracker <http://bugs.python.org/issue23967> ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue23967] Make inspect.signature expression evaluation more powerful
Peter McCormick added the comment: I missed the fact that Larry's patch obviates the need for the `builtins.` prefix, shortening the Argument Clinic parameter specification into: backlog: int(py_default="min(SOMAXCONN, 128)", c_default="Py_MIN(SOMAXCONN, 128)") -- ___ Python tracker <http://bugs.python.org/issue23967> ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue20175] Derby #6: Convert 50 sites to Argument Clinic across 8 files
Changes by Peter McCormick : -- nosy: +pdmccormick ___ Python tracker <http://bugs.python.org/issue20175> ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
[issue20175] Derby #6: Convert 50 sites to Argument Clinic across 8 files
Peter McCormick added the comment: The smallest of typo fixes. -- Added file: http://bugs.python.org/file39068/pdm-iofile_typo-v1.patch ___ Python tracker <http://bugs.python.org/issue20175> ___ ___ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
