From c31f4c0e66d2654c05d2dd7aee60e0d827408a1b Mon Sep 17 00:00:00 2001 From: Joey Hess Date: Mon, 17 Jun 2019 15:26:46 -0400 Subject: [PATCH] devblog --- ...__finally_background_checksum_verification.mdwn | 14 ++++++++++++++ 1 file changed, 14 insertions(+) create mode 100644 doc/devblog/day_594__finally_background_checksum_verification.mdwn diff --git a/doc/devblog/day_594__finally_background_checksum_verification.mdwn b/doc/devblog/day_594__finally_background_checksum_verification.mdwn new file mode 100644 index 0000000000..6a8845545a --- /dev/null +++ b/doc/devblog/day_594__finally_background_checksum_verification.mdwn @@ -0,0 +1,14 @@ +Finally got checksum verification running in a separate job pool from +downloads, to better keep bandwidth saturated. + +I had to resort to what felt like a bit of a hack, but I can't see a better +way to do it. Also, I got stuck for far too long on a STM deadlock bug. + +Interestingly, this means that -J1 now has a purpose, it's not the same as +no -J option. Instead, it lets one download and also one concurrent +checksum of the previous download run at the same time. + +It would be nice if -J1 could be the default.. One problem with that is +that it needs a unicode locale to work due to a limitation of +concurrent-output. Changing the concurrency method based on the locale does +not seem like a good idea.