| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
|
|
| |
* the storageserver didn't like getting the TERM signal both from
runserver (via killpg) and sentinel, so it would print a warning
to stderr. But the warning didn't end with a newline.
* the line-splitter used by config-sentinel didn't handle input
that didn't end with newline before EOF well, so the warning
was never handled and sent to the vespa.log file earlier.
* only print warning if too many signals seen (> 2), and end
it with a newline.
* don't ignore last byte of input if we can avoid it.
|
|
|
|
|
|
|
| |
* don't add extra newline after EOF received
* eof() should only return true after all input is consumed
* still need to handle EOF-without-newline, but
do it in a more conservative manner.
|
|
|
|
| |
* this was leading to really strange problems in system tests.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Fail test scripts on first error
* Some more test fixes
* Reverted prod_features_test fix
* Re-added prod_features_test.sh change matching that of PR #193
* More error checking
* Reverted set -e on failing tests
* Enable errexit in lowercasing test
|
| |
|
|
|
|
| |
* defaults to "yahoo" if VESPA_USER is not set
|
| |
|
|
|