OLAQ Article PDF errors on OJS-Test

[carakey]

For OLAQ journal issues migrated from Bepress (all issues prior to 25(4)), in the current 3.3 instance at ojs-test:

• Article PDFs don’t load in the viewer - there’s a red bar at the top with a “Missing PDF file” message.
• XML exporting does not work with the Native XML plugin - usually with a lot of XML validation errors and the whole section of the XML file apparently missing; occasionally fails with a 502 error instead.
• In the UI, the table of contents for an issue shows the correct articles, but clicking from there to go to any submission results in a blank white screen.
• Trying to access submissions from the submissions tab in the UI, the first page of archived submissions (where the articles are all in those most recent issues) loads fine, but trying to click to subsequent pages gets the “an unexpected error has occurred” message and nothing loads there.

Example: https://ojs-test.library.oregonstate.edu/olaq/article/view/vol25_iss2_2/1982

The reported XML validation errors when attempting to export an issue are:

Element '{http://pkp.sfu.ca}submission_file', attribute 'stage': [facet 'enumeration'] The value '' is not an element of the set {'public', 'submission', 'note', 'review_file', 'review_attachment', 'final', 'fair_copy', 'editor', 'copyedit', 'proof', 'production_ready', 'attachment', 'query', 'review_revision', 'dependent'}.

Element '{http://pkp.sfu.ca}file', attribute 'filesize': '' is not a valid value of the atomic type 'xs:int'.

(The above pair errors seem to occur twice for each submission item in the issue)

Element '{http://pkp.sfu.ca}issue': Missing child element(s). Expected is ( {http://pkp.sfu.ca}articles ).

FWIW:

• I have experienced the behavior both as an anonymous user and logged in as admin.
• I have not seen this problem in other journals.
• There does not seem to be any difference between browsers.
• The most obvious difference between this group of problematic issues versus working issues is the means of import - that these were fetched from a remote source rather than base-64 embedded in the XML import files.
  • The same issues export successfully with the Native XML Plugin in the production OJS 3.1 system, and those exports do include the element as well as filesize attributes with xs:int values and stage attributes with values from the enumerated set.
  • The three newest issues seem to work fine, as do the PDF issue galleys, as do the landing pages for affected issues and articles.

[wickr]

OLAQ Article PDFs imported came in with a file_stage ID set to 1 but it turns out that value is not used by OJS anymore. https://github.com/pkp/pkp-lib/blob/d55b811f5ac8a68b729a70288d35c4148411b49c/classes/submission/SubmissionFile.inc.php#L24

Since that was set incorrectly, the migration script for v3.3 failed moving PDF files to the correct path, and the DB table with file paths had an invalid path set: journals/1/articles/9//9-1-9-1-1-20200131.pdf

This was fixed manually, by moving the PDF files to the expected paths, and updating the DB tables that have the paths included. Updating DB file paths:

# single path update
UPDATE files SET path = 'journals/1/articles/994/submission/proof/994-1-994-1-1-20200131.pdf' WHERE file_id = 866;
# all paths update
UPDATE files SET path = REPLACE(path, '//', '/submission/proof/') WHERE file_id < 10;

Moving the PDF files ended up being a multi-step process, getting a list of all the article folder paths with public saved to a file, and then ultimately iterating over that list, and running

find . -name '*public' -exec sh -c 'mkdir -p submission; mv "$@" submission/proof' find-sh {} +

That made all of the PDFs display correctly, but the admin UI still wouldn't show more than 1 page of the submissions archive. Setting the file_stage values to 10 instead of 1 in the DB fixed the admin UI issue.

Updating DB submission_files table:

UPDATE submission_files SET file_stage = 10 WHERE submission_file_id < 1006;

Doing this before the v3.3 migration would almost certainly solve the above file path issues, since the migration code would recognize the file stage values.

v3.3 Migration error log messages:

PHP Notice:  Undefined offset: 1 in .../ojs-3.3.0-8/lib/pkp/classes/migration/upgrade/PKPv3_3_0UpgradeMigration.inc.php on line 609
A submission file was expected but not found at journals/1/articles/970//970-1-970-1-1-20200131.pdf.
A file assigned to the file stage 1 could not be migrated.

[carakey]

I spot checked several issues' PDFs and verified that exporting with Native plugin is working.

I found a discrepancy between the XSLT for original OLAQ files from Bepress (whose PDFs were broken in 3.3) versus XSLT for 2.x to 3.x (whose PDFs were working) - the submission_file stage attribute in the former being 'public' and latter being 'proof'.