Bug: bin/omero import --logs doesn't send log to QA

[bpindelski]

During testing of ​https://github.com/openmicroscopy/openmicroscopy/pull/1336, I came across the issue that specifying --logs and --report for a failing CLI import doesn't send any log files to QA. This was done on dev_4_4, so there was no per-image import log file as on develop (AFAIK). The log file was meant to be under /Users/bpindelski/omero/importer.log, but was missing.

Colin mentioned that this could have an impact also on Dropbox, as those CLI parameters are used there too.

[sbesson]

Retested during the review of ​https://github.com/openmicroscopy/openmicroscopy/pull/1339, same issue was found on the develop.

More thorough investigation shows the following call uploads the log and the image (see ​http://qa.openmicroscopy.org.uk/qa/feedback/7458/?token=ff7c054d05ed60bf68703fb0facf94ab)

bin/omero import ~/broken\ images\ from\ Andrew/created\ by\ Andrew\ -\ fails\ after\ upload/wrong-detector-settings-mk1.ome -- --report --email s.besson@dundee.ac.uk --upload --logs

while the following call does not upload any file (see ​http://qa.openmicroscopy.org.uk/qa/feedback/7456/?token=25a3822eb374bf11fe5a375a017a78c9)

bin/omero import ~/broken\ images\ from\ Andrew/created\ by\ Andrew\ -\ fails\ after\ upload/wrong-detector-settings-mk1.ome -- --report --email s.besson@dundee.ac.uk --logs

In both cases, "Sending log file..." is correctly displayed so the bug is likely to be in the ErrorHandler?.sendErrors() method

[jburel]

Review as part of Beta2 import work.

[jamoore]

Certainly just reproduced on develop. Couple of issues:

• passing --logs with --report does nothing. This is a usability issue.
• passing both of the above creates a QA item, but no logs are upload.

Quite likely the old HTTP code from the importer is now out of sync.

[jamoore]

Things as I understand them:

• Python args are properly being parsed to CommandLineImporter
• but ome.formats.importer.util.ErrorHandler.sendErrors() is not doing its job of uploading a client-side import file.

[bpindelski]

This might be more complicated than it looks at the surface. If a user specifies the usual required options and additionally --report --logs, the code path will reach blitz/src/ome/formats/importer/util/ErrorHandler.java. There we have

if (sendLogs)
{
    errorContainer.addFile(config.getLogFile());
}

config is an instance of ome.formats.importer.ImportConfig. The getLogFile() method calls getLogFile() on ome.formats.importer.util.IniFileLoader which finally has a hard-coded path pointing to <user_home_folder>/omero/importer.log. That explains why no file is found (as it's not being created in the first instance).

Two questions arise:

1. Should we keep using importer.log? If so, this code path will need to be checked and fixed so that the import process creates the necessary log file.
2. If we want per-file import logs (as it they exist on develop) to be uploaded, then the code path starting from ImportConfig is broken (i.e. relies on dev_4_4 logic) and need a bigger refactoring.

[jamoore]

1. importer.log should still be generated. Did it stop being generated when we moved to logback?
2. I'm not exactly sure what you mean by "per file". If you mean each of the import logs from the server-side filesets, then that will need to be an RFE for later.

[bpindelski]

Answering 2) first:
Yes, that's what I meant. Since there was no clear explanation, I thought that the --logs option would send import logs in the develop form (i.e. one per image file). As that isn't the case and (as you mentioned) requires an RFE, I shall debug what is happening with importer.log.

[jburel]

see ​https://github.com/openmicroscopy/openmicroscopy/pull/3289