GitLab

hy this merge is necessary,

New approach to protecting DBI handles from a timed-out disconnection of the server. It extends to methods like $dbc->do() or $dbc->select_*

New approach to protecting DBI::st(atement handlers) from a timed-out disconnection of the server. It extends to all methods that are prepare()d and then execute()d.

The "topup" mode was not error-safe because it the first run failed, the
second would almost certainly fail because of duplicated rows.
The solution is to add the --insertignore flag and to tweak the post-copy
checks. Even though the mysqldump options are now the same in "topup" and
"insertignore" modes, "topup" still means that all the source rows must be
in the target table (even though some could already be present when the
copy starts). "insertignore" means that we know that some rows could be
missing.

Don't need to consider the EHIVE_DROP_DB_TEST_URL variable, since we have a get_test_urls() in Utils::Test

* version/2.4:
  bugfix: "return_codes_2_branches" may be missing, in which case skip this part

Don't display DONE and EMPTY analyses in beekeeper output but still mentions how many of them there are.

This does not apply to the debug mode

* version/2.4:
  changes to allow no trailing slash in database-instance URLs (updated not to assume dbname is always defined)
  Explain why the module could not be loaded
  Cache the total hive load in HivePipeline

changes to allow no trailing slash in database-instance URLs (updated not to assume dbname is always defined)