aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorBoris Kolpackov <boris@codesynthesis.com>2019-10-23 15:36:00 +0200
committerBoris Kolpackov <boris@codesynthesis.com>2019-10-23 15:36:00 +0200
commit282361b57f66b59a50ed3d073b46c7189c9553ee (patch)
treec8c203cea468b52bc609b4ab9c2b9c86a575db48
parent1e577f5ff22e95786daedca5c326029baa59a39d (diff)
Add note in comment
-rw-r--r--bdep/init.cxx11
1 files changed, 9 insertions, 2 deletions
diff --git a/bdep/init.cxx b/bdep/init.cxx
index e6ddc1e..1abff60 100644
--- a/bdep/init.cxx
+++ b/bdep/init.cxx
@@ -103,12 +103,19 @@ namespace bdep
}
// Should we sync then commit the database or commit and then sync?
- // Either way we can end up with an incosistent state. Note, however,
+ // Either way we can end up with an inconsistent state. Note, however,
// that the state in the build configuration can in most cases be
// corrected with a retry (e.g., "upgrade" the package to the fixed
// version, etc) while if we think (from the database state) that the
// package has already been initialized, then there will be no way to
- // retry anything.
+ // retry anything (though it could probably be corrected with a sync or,
+ // failed that, deinit/init).
+ //
+ // However, there is a drawback to doing it this way: if we trigger an
+ // implicit sync (e.g., via a hook) of something that uses the same
+ // database, we will get the "database is used by another process"
+ // error. This can be worked around by disabling the implicit sync
+ // (BDEP_SYNC=0).
//
cmd_sync (o, prj, c, pkg_args, false /* implicit */);