Michael Peters [Thu, 20 Nov 2014 22:41:44 +0000 (17:41 -0500)]
LP#1394989: Make users_of_interest test for defined actor.usr.card values
If the open-ils.collections.users_of_interest API encounters a patron that
has a null "card" value in actor.usr the API will crash with an error like:
Received Exception:
Name: osrfMethodException
Status: *** Call to [open-ils.collections.users_of_interest.retrieve] failed for session [1416515351.333446.1416515351406], thread trace [1]:
Can't call method "barcode" on an undefined value at /usr/local/share/perl/5.14.2/OpenILS/Application/Collections.pm line 299.
This patches users_of_interest to provide "undef" in the event that there is
not a proper barcode associated with the user when retrieving results for the
API call.
Signed-off-by: Michael Peters <mpeters@emeralddata.net> Signed-off-by: Ben Shum <bshum@biblio.org>
Kathy Lussier [Fri, 7 Aug 2015 15:00:46 +0000 (11:00 -0400)]
Documentation: Add 2.8 docs for void on claims returned
Adding docs for void on claims returned feature. Also, adding a reference
to new "no billing activity on zero balance transaction" in one area of
circulation that was missed. Branch includes some other updates to old
information.
LP#1479953: Rename indexes to *_idx and add pgTAP test.
To make the new index names more consistent with existing indices,
I took the liberty of adding _idx to the new index names.
We also add a pgTAP test to verify that the new indices exist in
the database. The test would fail before running the upgrade
script and succeed after.
Signed-off-by: Jason Stephenson <jstephenson@mvlc.org>
Jeff Davis [Thu, 30 Jul 2015 20:12:35 +0000 (13:12 -0700)]
LP#1479953: Add indexes to vqbr foreign key references
Deleting a bib queue can be quite slow if it contains many queued
records. This is because each queued record needs to be deleted, and
this in turn affects many tables that contain unindexed foreign key
references to vandelay.queued_bib_record. Indexing those foreign keys
can speed up bib queue deletion substantially.
Signed-off-by: Jeff Davis <jdavis@sitka.bclibraries.ca> Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Jason Stephenson <jstephenson@mvlc.org>
Dan Wells [Wed, 29 Jul 2015 15:30:51 +0000 (11:30 -0400)]
LP 1198465: Load negative balance test transactions in load_all.sql
This commit integrates the SQL needed to setup negative balance
testing into the load_all.sql which loads the rest of the test data.
It also separates out the "reset" commands into a separate SQL file.
They are useful when doing repeated testing, but unnecessary and
may eventually require manual tweaking, so they are provided here as a
convenience only.
Signed-off-by: Dan Wells <dbw2@calvin.edu> Signed-off-by: Remington Steed <rjs7@calvin.edu>
Mike Rylander [Mon, 15 Jun 2015 19:32:20 +0000 (15:32 -0400)]
LP#1465385: Fix some syntax issues with make_release
Some versions of bash become confused when a unary = operator
is used in place of a binary == operator. Also, some variables
live without the protection of quoting, and thus further confuse
the shell when their contents are not as simple as originally
expected.
Here we adress those issues.
Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Ben Shum <bshum@biblio.org>
Terran McCanna [Fri, 15 May 2015 17:18:15 +0000 (10:18 -0700)]
LP1454879: Add Account Expiration Date to OPAC
Added Account Expiration Date to OPAC My Account display on
Account Summary page and Account Preferences page. This is in
response to patron requests to know when their accounts will
expire before the point at which they actually expire.
To test, log into My Account through the OPAC - a new table row
with account expiration date should appear on the main account
summary page, and another new row should appear on the main
account preferences page.
lp1466201 Disable Google Analytics in the staff client
As Galen Charlton noted in bug 1452883, "I suggest disabling Google Analytics
outright in the staff client; there's far too much potential for leaking patron
information." This branch aims to do so.
Test plan: Enable Google Analytics in config.tt2. Access the catalog in the
staff client. Click Debug to view the source code. Pre-patch, you'll see
the Google Analytics javascript within the head element. Post-patch, that
javascript will be gone. Viewing the source of a catalog page in a browser
should continue to show the Google Analytics javascript within the head element.
Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Ben Shum <bshum@biblio.org>
LP 1198465: Adapt some language in the negative balance branch
End users may see the term 'adjustment payment' and think that an actual
payment was made. Let's use 'account adjustment' instead. Also, remove any
references to credits in the description for the OU settings since it could
be confused with patron credits, which aren't prohibited by the code.
Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Ben Shum <bshum@biblio.org>
6. Restores Overdue Fines Appropriately, No Previous "Voids", Patron
Will Not Owe On Lost Item Return
7. Restores Overdue Fines Appropriately, No Previous "Voids", Patron
Will Still Owe On Lost Item Return
9. Restore Overdue Fines Appropriately, Previous Voids, Negative
Balance Allowed
13. Prohibit negative balances on lost materials bills ONLY
14. Prohibit negative balances on overdue bills ONLY
Note that test case 5 is omitted because it is a duplicate of case 2,
and case 11 is included but commented out because it is unclear how best
to handle the situation.
Signed-off-by: Remington Steed <rjs7@calvin.edu> Signed-off-by: Dan Wells <dbw2@calvin.edu> Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Ben Shum <bshum@biblio.org>
Dan Wells [Fri, 17 Jul 2015 20:45:16 +0000 (16:45 -0400)]
LP 1198465: Make conditional negative balances test sql re-runnable
Add some DELETEs and some more explicit setting of a few DB pieces such
that running the test SQL a second time will get everything back to
state where the tests will again proceed as expected (hopefully
successfully!).
Signed-off-by: Dan Wells <dbw2@calvin.edu> Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Ben Shum <bshum@biblio.org>
LP 1198465: Initial tests for conditional negative balances
This is a first commit of work-in-progress for testing the conditional
negative balances features. It covers 9 of the 14 test cases listed
here (as of today):
TODO:
- The test is currently an SQL setup file plus a Perl live test file.
One simple improvement would be to switch the setup process to
cstore calls within the Perl test file. This would be both more
contained and more robust.
- A second step to more advanced and useful tests would be to use
higher-level API calls to create portions of the setup rather than
doing everthing manually. However, some test conditions cannot be
reasonably setup with the normal API calls (e.g. bills of a specific
age, or bills using a legacy format no longer generated by current
code), so certain areas will likely always require direct
manipulation.
Signed-off-by: Daniel Wells <dbw2@calvin.edu> Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Ben Shum <bshum@biblio.org>
Dan Wells [Thu, 28 May 2015 17:54:26 +0000 (13:54 -0400)]
LP 1198465: Set restored overdue timestamp to time of last overdue
When we have our settings configured to generate new overdues on lost
item return, we start generation after the most recent overdue fine.
Because of this, we need the restored fine to be dated in the past,
which in turn allows the fine generator to apply catch-up fines as
expected.
Signed-off-by: Dan Wells <dbw2@calvin.edu> Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Ben Shum <bshum@biblio.org>
Remington Steed [Wed, 18 Jun 2014 13:34:13 +0000 (09:34 -0400)]
LP 1198465: Account for overdues voided the old way
The code for the particular case of "void"-overdues-for-lost is incomplete.
This is the one case in the new code where it now always adjusts rather
than voids, but the new restore code still needs to recognize cases where
the voiding happened under the old code.
Dan Wells [Tue, 1 Apr 2014 19:13:00 +0000 (15:13 -0400)]
LP 1198465: Delay creation of bill map for special bill handling
We were making the map, then deciding whether we wanted to void or
adjust. Since we only want the map if we adjust (and it creates havoc
when voiding), let's just get the bill IDs, then let the adjustment
code create the map it needs.
Signed-off-by: Dan Wells <dbw2@calvin.edu> Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Ben Shum <bshum@biblio.org>
- Replace ou_ancestor_setting() with ou_ancestor_setting_value() calls
This also fixed a bug where we were expecting just the setting, not
a HASH
- Reword interval checking
This fix is two part. First, we simplify the check to not require
the whole payment map. Second, we use this newfound simplicity to
push this check up into the gatekeeper functions, further clarifying
the code paths.
- make $note into $for_note for void_or_zero_bills_of_type()
Because the function can both void and adjust, we can't supply a
complete note, so let's just supply text of what the void/adjustment
is for.
Signed-off-by: Dan Wells <dbw2@calvin.edu> Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Ben Shum <bshum@biblio.org>
Dan Wells [Wed, 26 Feb 2014 16:12:21 +0000 (11:12 -0500)]
LP 1198465: Refactor logic into gatekeeper functions
The bulk of this commits take the logic from adjust_bills_to_zero() and
moves it up a layer into the "gatekeeper" void_or_zero* functions.
This move also allows us to simplify the logic, since some facts are
already known based on our function path.
Also:
- give void_or_zero_overdues() a new signature to better support
multiple options
- add new 'force_void' and 'force_zero' options to this function
- rename real_void_bills() to simply void_bills() (since there is no
other void_bills(), the "real" was redundant)
Signed-off-by: Dan Wells <dbw2@calvin.edu> Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Ben Shum <bshum@biblio.org>
Dan Wells [Tue, 25 Feb 2014 22:37:43 +0000 (17:37 -0500)]
LP 1198465: Restore voiding code, rename adjustment function
This commit restores the code we removed a few commits earlier, and
also renames "real_void_bills()" "adjust_bills_to_zero()" in order to
better delineate the split functionality.
Signed-off-by: Dan Wells <dbw2@calvin.edu> Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Ben Shum <bshum@biblio.org>
This new payment type complements the current void logic that flags
bills as voided.
This new payment type is needed because the current way that Evergreen
voids bills requires that all voids happen in the same increment as the
bills themselves. This prevents voiding of a partial bill or a bill
that has had a partial payment applied.
This commit also adds the org. unit setting types for the conditional
negative balances enhancements:
Finally, create a helper function for checking intervals along the way,
and a handy little subroutine to void all bills of a given type on a
circulation transaction.
* added some new locations and better groups (thanks gdunbar)
* applying locations to copies in a slightly more sane fashion
* added todo comment about possible future improvement
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Kathy Lussier <klussier@masslnc.org>
* Add/remove some locations
* Add some locations whose flags (circulate, etc.) vary along the org
pathway. Thanks to Thomas Berezansky for this suggestion.
* Ensure that copy location owner matches copy circ lib when applying
locations to copies.
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Kathy Lussier <klussier@masslnc.org>
Bill Erickson [Fri, 15 Mar 2013 18:32:18 +0000 (14:32 -0400)]
LP#1155771 Sample data : copy locations, location groups
Adds a pile of sample copy location and location groups to the sample
data in Open-ILS/tests/datasets/sql. Samples are loaded as part of
"load_all.sql"
Signed-off-by: Bill Erickson <berick@esilibrary.com> Signed-off-by: Kathy Lussier <klussier@masslnc.org>
Bill Erickson [Fri, 22 May 2015 20:07:29 +0000 (16:07 -0400)]
LP#1312308 remove deprecated open-ils.penalty service
open-ils.penalty service has not been used by Evergreen since before
version 2.0. Instead of just cleaning out the script configuration,
clean out the entire application.
Signed-off-by: Bill Erickson <berickxx@gmail.com> Signed-off-by: Jason Stephenson <jstephenson@mvlc.org>
Yamil Suarez [Fri, 19 Jun 2015 14:50:45 +0000 (10:50 -0400)]
LP#1465830: authority linker now ignores $e and $4 in bib name headings
presence of relator terms or codes in $e or $4 can prevent authority
headings linking
To test
-------
[1] Create an authority record with a name heading like this
100 $a Example, Natalia $d 1945-
[2] Create a bib record that contains the name and a relator
code or term, e.g.,
100 $a Example, Natalia $d 1945- $4edt
[3] Run authority_control_fields.pl. Note that the
heading in the bib is *NOT* linked, as evidence
by no $0 being added to the bib heading and no
rows added to authority.bib_linking for that bib
and authority.
[4] Apply the patch and re-run step 3. This time,
the link gets established:
100 $a Example, Natalia $d 1945- $4 edt $0 (CONS)xxx
Ben Shum [Tue, 3 Mar 2015 02:40:47 +0000 (21:40 -0500)]
Remove alpha and RC scripts for 2.3-2.4
These were in the repository from the testing phase of 2.3-2.4 during the
2.4 alpha/RC period. These script are not required and may be confusing if
run. Remove as unnecessary.
Signed-off-by: Ben Shum <bshum@biblio.org> Signed-off-by: Jason Stephenson <jstephenson@mvlc.org>
Mike Rylander [Thu, 11 Jun 2015 12:49:21 +0000 (08:49 -0400)]
LP#902255: Correct "ILS Event at check in" logic
In a classic case of "fix a bug to reveal another" it turn out that
there was a small logic bug hiding in the check in hold-capture code.
The result of this bug was that the wrong ILS Event object was being
returned when we ran into the double-scan hold-capture situation by
overwriting the correct one immediately after we construct it, which
in turn prevented the previous commit from working as designed.
We now return the correct ILS Event immediately so that we can
properly detect the situation. Two birds with one branch.
Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Jason Stephenson <jstephenson@mvlc.org>
Mike Rylander [Mon, 8 Jun 2015 15:11:04 +0000 (11:11 -0400)]
LP#902255: Protect against hold double-capture
When the staff client is in Fast Scan (async) mode, it is possible
to double-scan at checkin and cause two holds to be captured by the
same item. By creating a specially crafted unique index we ensure
that the database rejects this situation.
This protection would cause unhandled errors in the staff client, so
we also register a new string to be displayed when this particular
situation occurs, alerting staff to the possible double-scan of a
barcode at checkin IFF they are using async mode.
Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Jason Stephenson <jstephenson@mvlc.org>
Adam Bowling [Fri, 16 Jan 2015 22:18:24 +0000 (17:18 -0500)]
LP#1365661 Found patrons with same name not opening
This patch fixes the error, while registering a new patron
and clicking the link "Found {$0} patron(s) with the same
name" did not previously work in the staff client.
To reproduce the bug, register a new patron with the name of an
existing patron. The little widget with the message described
above will appear. The message is a hyperlink, but nothing
appears to happen if you click it. The JavaScript console will
contain an error ReferenceError:js2JSON is not defined.
After applying this patch, clicking the hyperlink does a search
for matching patrons as intended. The ReferenceError also
disappears.
Signed-off-by: Adam Bowling <abowling@emeralddata.net> Signed-off-by: Terran McCanna <tmccanna@georgialibraries.org> Signed-off-by: Jason Stephenson <jstephenson@mvlc.org>
Michael Peters [Tue, 9 Jun 2015 14:10:04 +0000 (10:10 -0400)]
LP#1154656 MARC Expert Search "Add Rows" adds duplicate row
This patch changes the behavior of the "Add Search Row" link
(JavaScript function from simple.js "addExpertRow()") so that
we are always cloning the initial "empty" row instead of
potentially cloning a row -- by way of cloneNode() -- with user
input in place, and causing duplicate search rows.
To reproduce the bug, go to Advanced Search, fill in all three
search fields with some text, click on Add Search Row. You wil
see that the Title search row along with the data in the text
box is duplicated and added to the search.
After this patch, the new row should be added with an empty
text box.
Signed-off-by: Michael Peters <mpeters@emeralddata.net> Signed-off-by: Rashma Kumaran <rkumaran@kcls.org> Signed-off-by: Jason Stephenson <jstephenson@mvlc.org>
Chris Sharp [Tue, 7 Apr 2015 00:17:30 +0000 (20:17 -0400)]
LP#1342227 - Allow EDI Ruby dependency installation on Ubuntu 14.04
Create a script for Ubuntu 14.04 ruby installation and installation
for EDI dependencies. Remove 'rcov', which has been deprecated for
Ruby 1.9.1 and install Bill Erickson's GIR improvements.
In the future we will need to remove this dependency altogether (see
bug 1373690), but for now, at least get it working on Ubuntu 14.04.
Signed-off-by: Chris Sharp <csharp@georgialibraries.org> Signed-off-by: Jason Stephenson <jstephenson@mvlc.org>
This commit borrows two screenshots from the Sitka Evergreen Training
Manual to replace one in the Alternate Serial Control View section.
Thanks Sitka!
Chris Sharp [Wed, 15 Apr 2015 18:05:23 +0000 (14:05 -0400)]
LP1444623: Remove "Safe" CPAN dependency from Debian/Ubuntu Makefile.install files
Following up on IRC discussion at http://irc.evergreen-ils.org/evergreen/2015-04-15#i_170386
it was agreed that it is no longer necessary to install Safe from CPAN
since it is included in the standard Perl package.
Signed-off-by: Chris Sharp <csharp@georgialibraries.org> Signed-off-by: Ben Shum <bshum@biblio.org>
Conflicts:
Open-ILS/src/extras/install/Makefile.debian-jessie
Terran McCanna [Wed, 10 Jun 2015 19:44:46 +0000 (15:44 -0400)]
LP#1396263: Fix for 'Return to Record' link breaks after texting
If user is already logged into OPAC when texting a call number from
the record screen, the 'Return to Record' link works properly. However,
if user is not logged in and is prompted to log in first, the 'Return to
Record' link actually returns the user to 'My Account.' This patch
corrects that problem and returns the user to the record regardless of
whether they were initially logged in or not.
Mike Rylander [Wed, 10 Jun 2015 18:17:37 +0000 (14:17 -0400)]
LP#1463973: Use dollar-quoting for report literals
When inserting string literal into SQL for Clark, we quote
apostrophes with a backslash. That worked fine until Postgres
changed its default for the standard_conforming_strings GUC to "on"
after which it no longer worked fine.
So, we will now use "dollar quoting" for string literals instead
in much the same way we do for our search queries.
Test plan
---------
Pre-patch: Try to run a report where a string value is required as
a filter parameter, and include an apostrophe in that string. The
report will fail with an SQL error.
Post-patch: Do the same, it will not fail.
Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Angela Kilsdonk <akilsdonk@esilibrary.com> Signed-off-by: Ben Shum <bshum@biblio.org>
Chris Sharp [Wed, 22 Apr 2015 13:54:11 +0000 (09:54 -0400)]
LP#1446860 - Correct mistaken logic in previous fix.
See https://bugs.launchpad.net/evergreen/+bug/1446860/comments/2
for reference. The logic was not enforcing the "if the editing
staff member doesn't have permission to edit a user in this
permission group" check. This corrects that.
Signed-off-by: Chris Sharp <csharp@georgialibraries.org> Signed-off-by: Ben Shum <bshum@biblio.org>
Chris Sharp [Tue, 21 Apr 2015 21:10:49 +0000 (17:10 -0400)]
LP#1446860 Staff were able to edit their own accounts.
This change reverses what appears to be a thinko in the
original programming. If the editing user is the same
as the user being edited, disable the Save button.
Signed-off-by: Chris Sharp <csharp@georgialibraries.org> Signed-off-by: Ben Shum <bshum@biblio.org>
Jeff Davis [Thu, 16 Apr 2015 15:58:17 +0000 (08:58 -0700)]
LP#1074096: Remove Bib Call Number from query type selector
Signed-off-by: Jeff Davis <jdavis@sitka.bclibraries.ca> Signed-off-by: Michael Peters <mpeters@emeralddata.net> Signed-off-by: Ben Shum <bshum@biblio.org>
Kathy Lussier [Thu, 28 May 2015 02:11:49 +0000 (22:11 -0400)]
LP#1403966: Remove publication-specific information from metarecord results page
The records that make up a matarecord search result typically contain different
publication information. Therefore, we should remove the master record's
specific publication information from that search results page.
Test plan
Prior to loading the patch, conduct a search on the Concerto dataset
for "brahms piano concerto no. 1" with "Group Formats and Editions" turned on.
In the brief view, you will see a call number and, in the more details view, you
will see publisher and physical description information. Run the same search
after loading the patch, and you will no longer see that information.
Signed-off-by: Kathy Lussier <klussier@masslnc.org> Signed-off-by: Andrea Neiman <aneiman@kent.lib.md.us> Signed-off-by: Ben Shum <bshum@biblio.org>
Blake Henderson [Mon, 8 Dec 2014 21:52:15 +0000 (15:52 -0600)]
LP#1194860: You have permission to override some of the failed holds. appearing when it should not for patrons in the OPAC Edit
A quick value check before assigning the override variable.
This should eliminate the OPAC showing the override option
for those who do not have the permission.
Josh Stompro [Sun, 31 May 2015 22:49:34 +0000 (17:49 -0500)]
LP#1456021 Enable apache modcgi during install for Debian Jessie
Apache mod cgi isn't enabled when installing apache2.4 on Debian Jessie,
this causes the /updates/manualupdate.html to not be processed and
to just return the raw code of that script since it uses mod cgi.
This adds 'cgi' to the list of apache modules to enable during the
dependency install.
Signed-off-by: Josh Stompro <stomproj@larl.org> Signed-off-by: Ben Shum <bshum@biblio.org>
Josh Stompro [Sun, 31 May 2015 22:19:58 +0000 (17:19 -0500)]
LP#1457094 Switch to package libtemplate-plugin-posix-perl for Debian Jessie
Debian Jessie includes a package for Template::Plugin::Posix that was
being installed via CPAN. This patch removes the CPAN install and adds
the Debian package.
Signed-off-by: Josh Stompro <stomproj@larl.org> Signed-off-by: Ben Shum <bshum@biblio.org>
Mike Rylander [Mon, 8 Jun 2015 21:30:52 +0000 (17:30 -0400)]
LP#1289486: Remove blocking constraint from audit table
The ACQ lineitem audit table was augmented with an fkey pointing
to the queued record that was related to the line item. That's
bad because there are also triggers which attempt to manage that
table. Hilarity ensues. We should not have an fkey on an audit
table, so, we now remove that.
In order to test this, first attempt to delete an ACQ queue and
have that fail. Apply the DB change, attempt again, and have it
succeed.
Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Dan Wells <dbw2@calvin.edu>
Mike Rylander [Fri, 22 May 2015 17:11:18 +0000 (13:11 -0400)]
LP#1449709: Always get a real hostname for the cache key
EGCatLoader overwrites $ctx->{hostname} when serving the staff client
version of TT-generated content. This is bad for us because we use
that as part of the TT Processor cache key. Instead, we'll just
always ask Apache, via $r.
Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Bill Erickson <berickxx@gmail.com>
Galen Charlton [Tue, 5 May 2015 22:15:39 +0000 (22:15 +0000)]
LP#1452366: allow EGWeb context loaders to have child_init actions
This patch adds the ability for EGWeb to call
an initialization function for context loaders
during the child_init phase of Apache backend
startup.
In particular, portions of the R/O object cache
for EGCatLoader are now initialized when a backend
starts up; testing indicates that this can shave
a couple seconds off the time it takes for a
backend to render a bib details page the first time.
Signed-off-by: Galen Charlton <gmc@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Bill Erickson <berickxx@gmail.com>
Galen Charlton [Wed, 6 May 2015 18:23:49 +0000 (18:23 +0000)]
LP#1452352: use fresh CStore editors when populating R/O cache
This change ensures that there's no longer a cached
editor with its original session locale grabbing
the wrong labels when a template invokes ctx.search_foo.
Signed-off-by: Galen Charlton <gmc@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Bill Erickson <berickxx@gmail.com>
Galen Charlton [Wed, 6 May 2015 13:50:32 +0000 (13:50 +0000)]
LP#1452352: fix generation of locale-specific RO object fetchers
The TPAC R/O object cache is now keyed by locale. This
fixes an issue where an OPAC session that starts in one
locale and gets switched to another could see (e.g.)
format labels for the original locale.
The issue fixed by this patch can be most readily reproduced
as follows:
[1] Enable both the en-US and fr-CA locales. Make sure
that there are some translations for record format
coded values in place as well.
[2] Run Apache in single-process mode
(e.g., ". /etc/apache2/envvars && apache2 -X")
[3] Visit a record details page in the en-US locale. Note
the label associated with the record format.
[4] Change the locale to fr-CA. Note that the label is
still the English version.
Signed-off-by: Galen Charlton <gmc@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Galen Charlton <gmc@esilibrary.com> Signed-off-by: Bill Erickson <berickxx@gmail.com>
This patch enables caching of compiled Template Toolkit
files on Evergreen web servers, which can provide a modest
improvement in the time it takes (say) TPAC to render
a page, particularly by a fresh Apache backend.
This is controlled by a new Apache virtualhost variable,
OILSWebCompiledTemplateCache, which can be set to a
directory on the webserver for storing compiled templates. This
is enabled by default for new installations.
This patch also adds OILSWebTemplateStatTTL, which can be
use to tweak the STAT_TTL Template Toolkit setting. Note this
bit works because of the caching of TT handlers added by
the previous patch.
Finally, this patch also fixes a bug where attempting to
disable OILSWebDebugTemplate would result in internal server errors;
it also sets the default value of this setting to false.
Note: the caching added by this patch and the previous one
mean that if a change to (say) TPAC templates is made on
the file system, it is now required to reload Apache to have
a guarantee that the change is visible, although one can also
wait the OILSWebTemplateStatTTL interval.
Signed-off-by: Galen Charlton <gmc@esilibrary.com> Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Galen Charlton <gmc@esilibrary.com> Signed-off-by: Bill Erickson <berickxx@gmail.com>
By persisting the Template Toolkit processor object per
vhost/locale/context-loader/xml-strictness instead
of creating a new one for every request, we allow TT
to maintain an in-memory cache of all the templates it
has compiled. This increases speed by as much as 80%.
In addition, we reduce the number of times the TPAC checks
for the existance of a template by remembering the state
of the first readability probe. This is most important
on systems where the templates are stored on NFS, where
stat(2) system calls can be particularly expensive.
We now pass the Apache object ($r) to the template processor
instead of relying on tied-STDOUT behaviour, because each
request gets a new tied file handle. After the first one,
it's the wrong one.
Signed-off-by: Mike Rylander <mrylander@gmail.com> Signed-off-by: Galen Charlton <gmc@esilibrary.com> Signed-off-by: Bill Erickson <berickxx@gmail.com>
projects / working / Evergreen.git / log