Movatterモバイル変換

bedevere-appbot mentioned this pull request

Show number of leaks in --huntrleaks#115720

Closed

vstinner reviewed

Lib/test/libregrtest/refleak.py

		symbol = 'X'
		if i == warmups:
		print(' ', end='', file=sys.stderr, flush=True)
		print(symbol, end='', file=sys.stderr, flush=True)

Copy link

Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others.Learn more.

To avoid any risk of side effect of that code, I suggest to add a get_symbol() function which would look into a pre-computed array for values <= 10. Something like:

symbols= ['.']+ [str(i)foriinrange(1,10)]+ ['X']defget_symbol(value):returnsymbols[min(max(value,0),10)]

vstinner reviewed

Lib/test/libregrtest/refleak.py

		else:
		symbol = 'X'
		if i == warmups:
		print(' ', end='', file=sys.stderr, flush=True)

Copy link

Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others.Learn more.

This is lovely, thanks :-)

Replace str() by tuple lookup

0dd589e

encukou commented

Lib/test/libregrtest/refleak.py Outdated

		if total_leaks <= 0:
		symbol = '.'
		elif total_leaks < 10:
		symbol = str(total_leaks)

Copy link

MemberAuthor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others.Learn more.

@vstinner instead of the function, maybe selecting a tuple of single-char strings would work better?

Suggested change

	symbol=str(total_leaks)
	symbol=(
	'.','1','2','3','4','5','6','7','8','9',
	)[total_leaks]

Copy link

Member

vstinner commentedFeb 20, 2024

The problem is how to display negative numbers:#115725

Copy link

MemberAuthor

encukou commentedFeb 20, 2024

Ah! Good point. Tomorrow I'll update this to make cases liketest_uuid looks nicer :)

Copy link

SIMBA111111 commentedFeb 20, 2024

veru good

vstinner reviewed

Feb 21, 2024

Reference count variations after getting__dict__,getattr, andgc.collect#115822

Copy link

Member

vstinner left a comment•
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others.Learn more.

I'm not fully convinced that it's a good idea to display intermediate results when we consider that there is no leak. As you saw with test_uuid, leaks can be subtle :-( There are worse cases such as#80741 which are really hard to reproduce and look like false alarms.

We already display values when we consider that there is a leak, and it's maybe enough?

If you have a lot of free time, you can changecheck_rc_deltas() heuristic to be stricter. But be careful of the rabbit hole, many people lost their mind in the Wonderland of reference leaks :-D

Show full result if it's not all zeros

f2a10a1

Copy link

MemberAuthor

encukou commentedFeb 21, 2024

What about showing the full result if it's suspicious (i.e. not all zeros), but using the current heuristic for failure (for now)?

12345:6789X4.6. 16.1test_uuid leaked [1, 6, -6, 1] references, sum=2 (this is fine)test_uuid leaked [0, 3, -3, 0] memory blocks, sum=0 (this is fine)== Tests result: SUCCESS ==

The heuristic is good (for now) for keeping CI green, but for bisecting I want to have all the info.

encukou added the 🔨 test-with-refleak-buildbotsTest PR w/ refleak buildbots; report in status section label

Feb 22, 2024

Copy link

bedevere-bot commentedFeb 22, 2024

🤖 New build scheduled with the buildbot fleet by@encukou for commitf2a10a1 🤖

If you want to schedule another build, you need to add the🔨 test-with-refleak-buildbots label again.

bedevere-bot removed the 🔨 test-with-refleak-buildbotsTest PR w/ refleak buildbots; report in status section label

Feb 22, 2024

Copy link

MemberAuthor

encukou commentedFeb 26, 2024

@vstinner Does this sound reasonable?
I'll merge today or tomorrow if there are no objections. We can make it better in future PRs :)

encukou merged commitaf5f9d6 intopython:main

Feb 27, 2024

bedevere-appbot removed the awaiting core review label

Feb 27, 2024

encukou deleted the huntrleaks-digits branch

February 27, 2024 08:51

encukou mentioned this pull request

Feb 27, 2024

Open

woodruffw pushed a commit to woodruffw-forks/cpython that referenced this pull request

Mar 4, 2024

pythongh-115720: Show number of leaks in huntrleaks progress reports (p…

f668718

…ythonGH-115726)Instead of showing a dot for each iteration, show:- '.' for zero (on negative) leaks- number of leaks for 1-9- 'X' if there are more leaksThis allows more rapid iteration: when bisecting, I don't needto wait for the final report to see if the test still leaks.Also, show the full result if there are any non-zero entries.This shows negative entries, for the unfortunate cases wherea reference is created and cleaned up in different runs.Test *failure* is still determined by the existing heuristic.

adorilson pushed a commit to adorilson/cpython that referenced this pull request

Mar 25, 2024

pythongh-115720: Show number of leaks in huntrleaks progress reports (p…

9063b8f

…ythonGH-115726)Instead of showing a dot for each iteration, show:- '.' for zero (on negative) leaks- number of leaks for 1-9- 'X' if there are more leaksThis allows more rapid iteration: when bisecting, I don't needto wait for the final report to see if the test still leaks.Also, show the full result if there are any non-zero entries.This shows negative entries, for the unfortunate cases wherea reference is created and cleaned up in different runs.Test *failure* is still determined by the existing heuristic.

vstinner pushed a commit to vstinner/cpython that referenced this pull request

pythongh-115720: Show number of leaks in huntrleaks progress reports (p…

2633b09

…ythonGH-115726)Instead of showing a dot for each iteration, show:- '.' for zero (on negative) leaks- number of leaks for 1-9- 'X' if there are more leaksThis allows more rapid iteration: when bisecting, I don't needto wait for the final report to see if the test still leaks.Also, show the full result if there are any non-zero entries.This shows negative entries, for the unfortunate cases wherea reference is created and cleaned up in different runs.Test *failure* is still determined by the existing heuristic.(cherry picked from commitaf5f9d6)

vstinner pushed a commit to vstinner/cpython that referenced this pull request

pythongh-115720: Show number of leaks in huntrleaks progress reports (p…

01436c5

…ythonGH-115726)Instead of showing a dot for each iteration, show:- '.' for zero (on negative) leaks- number of leaks for 1-9- 'X' if there are more leaksThis allows more rapid iteration: when bisecting, I don't needto wait for the final report to see if the test still leaks.Also, show the full result if there are any non-zero entries.This shows negative entries, for the unfortunate cases wherea reference is created and cleaned up in different runs.Test *failure* is still determined by the existing heuristic.(cherry picked from commitaf5f9d6)

vstinner mentioned this pull request

[3.12] gh-83434: Sync libregrtest and test_regrtest with the main branch#117250

Merged

vstinner added a commit that referenced this pull request

[3.12]gh-83434: Sync libregrtest and test_regrtest with the main bra…

477ef90

…nch (#117250)*gh-115122: Add --bisect option to regrtest (#115123)* test.bisect_cmd now exit with code 0 on success, and code 1 on  failure. Before, it was the opposite.* test.bisect_cmd now runs the test worker process with  -X faulthandler.* regrtest RunTests: Add create_python_cmd() and bisect_cmd()  methods.(cherry picked from commit1e5719a)*gh-115720: Show number of leaks in huntrleaks progress reports (GH-115726)Instead of showing a dot for each iteration, show:- '.' for zero (on negative) leaks- number of leaks for 1-9- 'X' if there are more leaksThis allows more rapid iteration: when bisecting, I don't needto wait for the final report to see if the test still leaks.Also, show the full result if there are any non-zero entries.This shows negative entries, for the unfortunate cases wherea reference is created and cleaned up in different runs.Test *failure* is still determined by the existing heuristic.(cherry picked from commitaf5f9d6)*gh-83434: Disable XML in regrtest when -R option is used (#117232)(cherry picked from commitd52bdfb)---------Co-authored-by: Petr Viktorin <encukou@gmail.com>

miss-islington pushed a commit to miss-islington/cpython that referenced this pull request

[3.12]pythongh-83434: Sync libregrtest and test_regrtest with the ma…

23a5deb

…in branch (pythonGH-117250)*pythongh-115122: Add --bisect option to regrtest (pythonGH-115123)* test.bisect_cmd now exit with code 0 on success, and code 1 on  failure. Before, it was the opposite.* test.bisect_cmd now runs the test worker process with  -X faulthandler.* regrtest RunTests: Add create_python_cmd() and bisect_cmd()  methods.(cherry picked from commit1e5719a)*pythongh-115720: Show number of leaks in huntrleaks progress reports (pythonGH-115726)Instead of showing a dot for each iteration, show:- '.' for zero (on negative) leaks- number of leaks for 1-9- 'X' if there are more leaksThis allows more rapid iteration: when bisecting, I don't needto wait for the final report to see if the test still leaks.Also, show the full result if there are any non-zero entries.This shows negative entries, for the unfortunate cases wherea reference is created and cleaned up in different runs.Test *failure* is still determined by the existing heuristic.(cherry picked from commitaf5f9d6)*pythongh-83434: Disable XML in regrtest when -R option is used (pythonGH-117232)(cherry picked from commitd52bdfb)---------(cherry picked from commit477ef90)Co-authored-by: Victor Stinner <vstinner@python.org>Co-authored-by: Petr Viktorin <encukou@gmail.com>

miss-islington mentioned this pull request

[3.11] [3.12] gh-83434: Sync libregrtest and test_regrtest with the main branch (GH-117250)#117251

Merged

vstinner added a commit that referenced this pull request