sapling/eden/fs/store/hg
Wez Furlong 65be9676eb fix for falling back to flat manifest import
Summary:
akushner reported this error situation with full details in this paste:

{P58193090}

(see also comments on the paste which include the edenfs.log)

Summarizing here, what appears to be happening is that there is a local commit
for which there is no local tree manifest data.  We try to load this from the
local treepacks and fail, so we fall back to asking the server, which fails.

In this case we want to fall back to performing a flat manifest import but
since the error is a generic runtime_error it doesn't get caught by the
code that decides to fallback on line 639.

Let's assume that an error reading the chunk header for the prefetchtrees
request is really a MissingKeyError so that we can attempt the fallback.

Reviewed By: simpkins

Differential Revision: D5898439

fbshipit-source-id: ea4f3f3ef22ee72c888c592a93c5c8b9269f39b0
2017-09-29 11:43:17 -07:00
..
test clean up treemanifest initialization in HgImporter 2017-08-17 16:13:00 -07:00
hg_import_helper.py add additional info to the hg_import_helper "started" response 2017-08-04 18:46:54 -07:00
HgBackingStore.cpp use rocksdb column families to separate key spaces 2017-09-07 14:50:42 -07:00
HgBackingStore.h use AbsolutePathPiece for repository path arguments 2017-06-23 15:23:32 -07:00
HgImporter.cpp fix for falling back to flat manifest import 2017-09-29 11:43:17 -07:00
HgImporter.h move LocalStore::PutXXX methods to a LocalStore::WriteBatch class 2017-09-28 20:53:40 -07:00
HgManifestImporter.cpp move LocalStore::PutXXX methods to a LocalStore::WriteBatch class 2017-09-28 20:53:40 -07:00
HgManifestImporter.h move LocalStore::PutXXX methods to a LocalStore::WriteBatch class 2017-09-28 20:53:40 -07:00
TARGETS use the newer fb-hgext code from scm/hgext 2017-08-15 12:15:06 -07:00
tester.cpp Remove no-longer-needed @manual for rocksdb 2017-09-01 19:31:46 -07:00