On systems where any new mount point may be probed by active daemons (such as gvfs-udisks2-volume-monitor), our test scripts may not be the first processes to trigger projection on the top-level directory of their test mount points; thus, the t300-args-initial.t test script would fail.
Therefore, in order to reliably test that the --initial mount option is respected, we add a new --log option to capture a record of successful projection requests, and then parse those logs (after ensuring at least one directory projection request occurs) to validate the functionality of the --initial flag.
Having the log file capability in the library allows us to fix a long-standing problem whereby event handler failures in send_event() were logged directly to stderr, which is not really appropriate for a library that will be included into larger applications. However, changing this to output event handler errors into the log file (if any) then breaks the t201-event-err.t test script, as it expects to capture the output that was previously sent to stderr and compare it to a generated set of expected error messages.
Hence, we add support for the --initial and --log options to our common test library, which allows us to continue to use the test_handlers helper program in the t201 script, as well as the test_simple helper program in t300 script, and thus removes the need for a dedicated test_args helper.
Having three potential logs (from stderr, stdout, and the optional log generated by --log) to juggle in our test programs, we revise some of the internal terminology and comments to help clarify which is which.
Finally, we can replace all the calls to fprintf() in our library with a log_printf() which logs to the specified log file (if any) and/or to stderr, according to an enum option specified by the caller and whether a log file exists and is open. This also makes the individual log message format strings more readable, as we can eliminate common elements like the projfs: prefix and trailing newline from each.
On systems where any new mount point may be probed by active daemons (such as
gvfs-udisks2-volume-monitor), our test scripts may not be the first processes to trigger projection on the top-level directory of their test mount points; thus, thet300-args-initial.ttest script would fail.Therefore, in order to reliably test that the
--initialmount option is respected, we add a new--logoption to capture a record of successful projection requests, and then parse those logs (after ensuring at least one directory projection request occurs) to validate the functionality of the--initialflag.Having the log file capability in the library allows us to fix a long-standing problem whereby event handler failures in
send_event()were logged directly tostderr, which is not really appropriate for a library that will be included into larger applications. However, changing this to output event handler errors into the log file (if any) then breaks thet201-event-err.ttest script, as it expects to capture the output that was previously sent tostderrand compare it to a generated set of expected error messages.Hence, we add support for the
--initialand--logoptions to our common test library, which allows us to continue to use thetest_handlershelper program in thet201script, as well as thetest_simplehelper program int300script, and thus removes the need for a dedicatedtest_argshelper.Having three potential logs (from
stderr,stdout, and the optional log generated by--log) to juggle in our test programs, we revise some of the internal terminology and comments to help clarify which is which.Finally, we can replace all the calls to
fprintf()in our library with alog_printf()which logs to the specified log file (if any) and/or tostderr, according to anenumoption specified by the caller and whether a log file exists and is open. This also makes the individual log message format strings more readable, as we can eliminate common elements like theprojfs:prefix and trailing newline from each.