With several promisor remotes, Git tries them in the order they appear in the configuration. Nelly Yopa's patch adds a promisor.fetchOrder variable holding a comma-separated list of remote names. Listed remotes are tried first, in order. Unlisted ones follow in their config order. The extensions.partialClone remote is still moved to the end.

Tanga Joris tested the first version on master in a partial clone. Setting promisor.fetchOrder produced "BUG: string-list.c:340: string_list_split_in_place() called with strdup_strings" on 'git show HEAD:f'. Joris traced it to a list created with STRING_LIST_INIT_DUP but then used with string_list_split_in_place_f(). Joris also noted that names were appended to the same list each time the variable was seen and never cleared.

Yopa says a v2 addresses these points. The list now lives in struct promisor_remote_config rather than a file-static, so it is per repository and cleared with the other promisor config. It is initialized with string_list_init_nodup(). The documentation entry was added to Documentation/config/promisor.adoc and the whitespace was fixed. The linked-list insertion sort is gone: try_promisor_remotes() now walks the list twice, first in fetchOrder order.

Junio C Hamano asked Christian, whom he described as the go-to person for promisor remotes, how this would mesh with the work letting the server tell clients which promisors to use first. Kaartic Sivaraam pointed to SubmittingPatches on describing changes. He said the patch seems to attempt part of the project without going through the RFC step, and asked whether Yopa had misunderstood what a microproject is. No answer on either point appears in the excerpts.