From 156adb9af66a17c69e6d402d1e3d3bd477af72be Mon Sep 17 00:00:00 2001 From: Joey Hess Date: Mon, 2 Oct 2017 11:56:11 -0400 Subject: forgot to add comment --- ...ent_4_a4bbb2efcf75167ac938b00128b340fd._comment | 22 ++++++++++++++++++++++ 1 file changed, 22 insertions(+) create mode 100644 doc/bugs/git_annex_test_fails/comment_4_a4bbb2efcf75167ac938b00128b340fd._comment diff --git a/doc/bugs/git_annex_test_fails/comment_4_a4bbb2efcf75167ac938b00128b340fd._comment b/doc/bugs/git_annex_test_fails/comment_4_a4bbb2efcf75167ac938b00128b340fd._comment new file mode 100644 index 000000000..7c132a808 --- /dev/null +++ b/doc/bugs/git_annex_test_fails/comment_4_a4bbb2efcf75167ac938b00128b340fd._comment @@ -0,0 +1,22 @@ +[[!comment format=mdwn + username="joey" + subject="""comment 4""" + date="2017-09-30T01:24:35Z" + content=""" +Built with ghc 8.2.1 on linux, I am seeing significantly more sub-processes +and open files. Up to 170 sub-processes and 350 open files. + +Most of the open files are pipes. +Most of the sub-processes are git cat-file --batch, and these are +definitely lingering from test repos it's no longer using. + +Kind of looks like the annex state created by Test.annexeval +is not getting garbage collected, and so the handles it has open are +not getting closed. Hmm, I don't normally rely on GC to close handles, +and am a bit surprised that ever worked, since the handles are pipes to +processes that have no reason to exit unless the handles are closed. + +Made it explicitly stop the co-processes, and that seems to have fixed it. +Indeed, the number of git subprocesses dropped to mostly below 10 since +they get cleaned up without waiting for GC now. +"""]] -- cgit v1.2.3