LP#1612771: Add chunking support to JS implementation

Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1612771: implement C max_chunk_size server support

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1612771: set Perl / C max_chunk_size default sizes

default max bundle size == 25K
default max chunk size  == 50K

Note with Ejabberd using 65536 as the default max stanza size, these
new OpenSRF defaults mean that all messages will fit the default
message size constraints -- i.e. no more need to raise the
max_stanza_size.

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1612771: C support for receiving chunked responses

* client parsing
* consistent w/ Perl, we now have "bundling" and "chunking"

Signed-off-by: Bill Erickson <berick@esilibrary.com>

LP#1612771: Perl max_chunk_size additions

* Added missing max_chunk_size method to AppSession
* Copy API max_chunk_size value into the handler AppRequest
* Fix error where no-chunking resulted in empty responses

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1612771: bundling and chunking

This patch is first in a series of patches that provides the following
features:

* OpenSRF message bundling -- Pack multiple OpenSRF messages together
in a single XMPP envelope, as long as we believe more messages will be
sent in the future and we are below some threshold of combined message size.
The default for that threshold is 25Kb.

 * OpenSRF message chunking -- Break up large OpenSRF messages across
multiple XMPP envelopes. This is implemented with a new OpenSRF message type.
C, Perl, and Javascript libraries are taught how to reconstruct chunked
messages. The default chunking threshold is 50Kb, just a bit below the default
ejabberd max stanza size of 64Kb.

This patch in particular renames "chunking" to "bundling", then
implements message splitting ("chunking") in Perl using two new
oilsResult subclasses

Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1603708: Remove support for Ubuntu 12.04 Precise

With support for Ubuntu 16.04 Xenial in place, remove the oldest LTS, which is
Ubuntu 12.04 Precise.

Signed-off-by: Ben Shum <ben@evergreener.net>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

Docs: Change 14.04 to Trusty in README

For consistency, change references from numbered "Ubuntu 14.04" to "Ubuntu Trusty"

Signed-off-by: Ben Shum <ben@evergreener.net>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

Docs: Add Xenial references in the websocket setup instructions

We were missing Xenial references for the websocket setup instructions.

They are the same as Trusty, so group together for now.

Also, change references to "Ubuntu 16.04" to read "Ubuntu Xenial" too.

Signed-off-by: Ben Shum <ben@evergreener.net>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1473479 Syslog configuration adoption

When an OpenSRF client is run with syslog enabled and with the
OSRF_ADOPT_SYSLOG environment variable set to a true value, no attempt
is made to modify the syslog configuration, including no calls to
openlog()/closelog() and no modification of the syslog facility when
calling syslog().

Signed-off-by: Bill Erickson <berickxx@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1631520: configure install location of Perl modules

Add --with-perlbase configure option to specify
an alternative location for installing the Perl modules. This
can be useful for setups that want to run the Perl modules
from a shared filesystem or environments that need to run
multiple versions of OpenSRF simultaneously.

Users of --with-perlbase are responsible for ensuring that
PERL5LIB is set appropriately.

To test
-------
[1] Use --with-perlbase during the configure step, e.g.,

    ./configure --perl-base /tmp/perl

[2] Run make; make check; sudo make install
[3] Verify that the Perl modules are installed under
    /tmp/perl.
[4] Make a change to a Perl source file, then
    go to src/perl, then run sudo ./Build install. Verify
    that it remembers the --with-perlbase directory
    and installs the updated module there.

Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1585041: Move debian_sys_config target for Debian distributions

Similar to how things were reordered for Ubuntu, let's move debian_sys_config
for Debian distributions.

Signed-off-by: Ben Shum <ben@evergreener.net>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1485371: Release notes for TZ handling in OpenSRF

Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1485371: Use client-supplied TZ

Currently, there is no protocol-level mechanism for passing the client's
desired timezone to the server. In much the same way we pass the locale,
we can let the server know what timezone it should use when interpreting
time stamps.

To do this we:

 * Teach perl server code to live in the client TZ, if supplied
 * Teach perl client code to send the current $ENV{TZ}
 * Teach javascript library to include client TZ in
   gateway/translator/websocket communication
 * Teach C code to pull the incoming TZ and apply it to outgoing messages
 * Teach srfsh to pull TZ from the environment and pass it with requests

Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1551090: Update README for Ubuntu 16.04 (Xenial Xerus).

Add installation steps for Ubuntu 16.04.

Signed-off-by: Jason Stephenson <jason@sigio.com>
Signed-off-by: Ben Shum <ben@evergreener.net>

LP#1551090: Enable mod_perl2 on Ubuntu 16.04 (Xenial Xerus).

Ubuntu Xenial Xerus does not automatically enable mod_perl2 for
Apache2 when the package is installed, so we enable it via the
Makefile.install.

Signed-off-by: Jason Stephenson <jason@sigio.com>
Signed-off-by: Ben Shum <ben@evergreener.net>

LP#1551090: Adding apache2-dev dependency to xenial and fixing whitespace

Signed-off-by: Chris Sharp <csharp@georgialibraries.org>
Signed-off-by: Jason Stephenson <jason@sigio.com>
Signed-off-by: Ben Shum <ben@evergreener.net>
Conflicts:
src/extras/Makefile.install

LP#1551090: Since we move apache to the "install_extra_debs*" targets,

we need to move debian_sys_config further down to compensate.

Signed-off-by: Chris Sharp <csharp@georgialibraries.org>
Signed-off-by: Jason Stephenson <jason@sigio.com>
Signed-off-by: Ben Shum <ben@evergreener.net>

LP#1551090: Adding necessary connectivity for xenial deb installation.

Signed-off-by: Chris Sharp <csharp@georgialibraries.org>
Signed-off-by: Jason Stephenson <jason@sigio.com>
Signed-off-by: Ben Shum <ben@evergreener.net>

LP#1551090: Adding apache2 package to Makefile.install deb list.

Signed-off-by: Chris Sharp <csharp@georgialibraries.org>
Signed-off-by: Jason Stephenson <jason@sigio.com>
Signed-off-by: Ben Shum <ben@evergreener.net>

LP#1551090: Updating Makefile to accommodate ubuntu-xenial target.

Removing some redundancy in the extra debs lists.

Signed-off-by: Chris Sharp <csharp@georgialibraries.org>
Signed-off-by: Jason Stephenson <jason@sigio.com>
Signed-off-by: Ben Shum <ben@evergreener.net>
Conflicts:
src/extras/Makefile.install

LP#1559121: remove Debian Squeeze support

Now that Debian Squeeze's LTS (long-term support) period
has ended, Evergreen no longer offers community support
for that distribution. This patch removes references
to Squeeze from the installation scripts and documentation.

To test:

[1] Verify that Debian Squeeze is no longer referenced in
    the installation documentation.

Signed-off-by: Galen Charlton <gmc@esilibrary.com>
Signed-off-by: Ben Shum <ben@evergreener.net>

LP#1350457: add test case for perl2JSONObject change

Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1350457: Pass caller's session to subrequests called via method_lookup

In the process of looking up a method for an internal subrequest, we lose
session info. This is a problem when the subrequest makes a remote request,
because then the subrequest can't look up the proper locale, among other
things. The forthcoming branch passes the caller's session to the subrequest.

This patch also teaches OpenSRF object registration how to strip certain
object members -- in particular, the session -- so that introspection
continues to work.

Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1474507: fix interval_to_seconds for weeks and seconds

This patch fixes an issue where OpenSRF::Utils::interval_to_seconds()
was not recognizing intervals expressed as seconds or weeks.

Signed-off-by: Jason Etheridge <jason@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1474507: tests for interval_to_seconds

Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1494486: Limit damage caused by dropped drone XMPP sockets

It is apparently possible for drones to get into a state where their XMPP
socket is closed but they don't notice. This is bad because the drone can
continue to receive requests from its listener but can no longer respond
to them. To limit the pain this can cause, we should kill the drone as soon
as we notice this condition.

To avoid overhead, this commit notices when the socket returns an error (or
raises a signal, in Perl) upon write, and exits immediately.  One message
will be lost, but the drone will no longer be a black hole that does nothing
but absorb requests it can never fill.

To test
-------
[1] Start an OpenSRF stack and look for a drone process.
[2] Use lsof to identify which socket that drone is using
    to talk to XMPP.
[3] Use gdb to attach to the process and close the socket, e.g.,

    $ gdb -p $PID
    (gdb) p close(11) # or whatever the socket number was
    (gdb) c

[4] Use srfsh to make requests of that service. Eventually, one
    of them will hit the drone.
[5] Sans patch, the request will get handled by the drone, but
    the results will never get sent, and the drone will remain
    available to handle other requests.
[6] With the patch, the drone will exit when it discovers that it
    can no longer write to the XMPP socket.

Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1152272: do not log batch invocation of srfsh in .srfsh_history

When running srfsh to process a script, do not log
the requests in .srfsh_history.

To test
-------
[1] Run the following command:

    echo 'request opensrf.math opensrf.system.echo "foo"' | srfsh

[2] Note that the last line of ~/.srfsh_history should be
    'request opensrf.math opensrf.system.echo "foo"'

[3] Apply the patch, compile, and run

    echo 'request opensrf.math opensrf.system.echo "bar"' | srfsh

[4] This time, no additional line should have been added to
    ~/.srfsh_history

[5] Another variation to try:

    srfsh test.srfsh

Signed-off-by: Galen Charlton <gmc@esilibrary.com>
Signed-off-by: Mike Rylander <mrylander@gmail.com>

LP#1461625: ensure srfsh doesn't close STDOUT prematurely

Ensure that when running srfsh in non-interactive mode
that reads commands directly from a file, (i.e.,
"srfsh script.srfsh" or as a shebang script), it does
not close STDOUT after handling the first request.

To test
-------
[1] Create a srfsh script containing:

    request opensrf.math opensrf.system.echo "foo"
    request opensrf.math opensrf.system.echo "bar"

[2] Run "srfsh script.srfsh". Note that only the
    results of the first echo request are output.
[3] Apply the patch and recompile, then run
    "srfsh script.srfsh" again.  This time, the
    output of both requests is displayed.

Signed-off-by: Galen Charlton <gmc@esilibrary.com>
Signed-off-by: Mike Rylander <mrylander@gmail.com>

LP#1409055 Support specific protocols for OpenSRF gateway requests

If the user passes in a gateway hostname that contains a specific HTTP or HTTPS
protocol, let's just use that.

Signed-off-by: Dan Scott <dscott@laurentian.ca>
Signed-off-by: Bill Erickson <berickxx@gmail.com>

LP#1436047: make srfsh --safe act as if "! command" doesn't exist

This patch make srfsh treat attempting to run an external
command via "! command" as a parsing error if --safe is
supplied.  It also suppress mention of "! commands" from
the internal help.

Signed-off-by: Galen Charlton <gmc@esilibrary.com>
Signed-off-by: Mike Rylander <mrylander@gmail.com>

LP#1436047: Allow disabling of "bang commands" in srfsh

srfsh has the ability to execute commands via system() calls using
the common "!command" syntax.  This is very useful, but it would
be nice to be able to turn that functionality off in some cases.

This branch adds argument parsing to detect a new '--safe' command
line parameter, which disables the "!command" syntax.

Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

Docs: Fix mailing list link for help in README

Signed-off-by: Ben Shum <bshum@biblio.org>

LP#1445503 - Updated Ejabberd setup steps for Ejabberd 14.x for Debian Jessie

Signed-off-by: Josh Stompro <github@stompro.org>
Signed-off-by: Ben Shum <bshum@biblio.org>

Docs: Emphasize variables and paths consistently in README

Some variables and paths were not emphasized when the instructions for
websockets was added. Maintain consistency in the document and emphasize
them.

Signed-off-by: Ben Shum <bshum@biblio.org>

Docs: Add [source, bash] to code blocks that were not defined in README

For consistency, add [source, bash] to all the blocks of code or script
commands in the README

Signed-off-by: Ben Shum <bshum@biblio.org>

Docs: Keep all source syntax consistent in README

Change occurrences of [source,bash] to [source, bash] for consistency
throughout README.

Signed-off-by: Ben Shum <bshum@biblio.org>

LP#1418613 per-tab websocket send() JS thinko repair

Avoid referencing variable defined somewhere outside the send_ws()
function.  Doing so happened to result in re-sending the same message
twice in some cases.

Signed-off-by: Bill Erickson <berickxx@gmail.com>
Signed-off-by: Mike Rylander <mrylander@gmail.com>

LP#1383828 Jessie/Trusty installer Apache "mpm" repairs

Disable the default Apache mpm_event module and enable mpm_prefork for
Debian Jessie and Ubuntu Trusty.

Signed-off-by: Bill Erickson <berickxx@gmail.com>
Signed-off-by: Ben Shum <bshum@biblio.org>

LP#1391248: Fix NameVirtualHost warnings in websockets apache2.conf

For the websockets configuration, the sample apache2.conf for Apache 2.4 (i.e.
Ubuntu 14.04, etc.) contains NameVirtualHost entries that are no longer
needed.

When starting apache2-websockets, you may see warnings like:

AH00548: NameVirtualHost has no effect and will be removed in the next
release /etc/apache2-websockets/apache2.conf:53

Remove the NameVirtualHost entries and these warnings go away.

Signed-off-by: Ben Shum <bshum@biblio.org>
Signed-off-by: Bill Erickson <berickxx@gmail.com>

LP#1369169: Remove hardcoded paths for OpenSRF directory

Signed-off-by: Ben Shum <bshum@biblio.org>
Signed-off-by: Bill Erickson <berickxx@gmail.com>

LP#1369169: Mention the requirement for valid SSL certificate

The apache2-websockets instance will not start without a valid SSL certificate
in /etc/apache2/ssl.  Include a mention of this in the README with the extra
stipulation that it is still possible to use a self-signed SSL certificate for
testing purposes, but this is not recommended for live installations.

Signed-off-by: Ben Shum <bshum@biblio.org>
Signed-off-by: Bill Erickson <berickxx@gmail.com>

LP#1369169: Add websockets section to the OpenSRF README

Remove the separate README.websockets and move the contents into the primary
OpenSRF README document so that all steps are in one place.

Additional edits to the websockets instructions to detail differences made
between Ubuntu 14.04 Trusty and Debian / Ubuntu 12.04 Precise. More edits
may be necessary for Debian Jessie later?

Also, create a separate config file for Apache 2.4 that is copied into place
for Ubuntu Trusty and potentially other systems that will need it.

Signed-off-by: Ben Shum <bshum@biblio.org>
Signed-off-by: Bill Erickson <berickxx@gmail.com>

LP#1369169: Add Apache 2.4 specific configuration file for websockets

For Apache 2.4, there were some necessary modifications for running the
websockets code properly. Similar to how we do things in Evergreen, we
added a new directory for apache_24 which contains the modified apache2.conf
file.

Signed-off-by: Ben Shum <bshum@biblio.org>
Signed-off-by: Bill Erickson <berickxx@gmail.com>

LP#1369169: Remove trailing whitespace from apache2.conf

For some reason, there was a lot of trailing whitespace in this apache2.conf
file for websockets. Removing them.

Signed-off-by: Ben Shum <bshum@biblio.org>
Signed-off-by: Bill Erickson <berickxx@gmail.com>

LP#1369169: Standardize styling of root linux account for README

Signed-off-by: Ben Shum <bshum@biblio.org>
Signed-off-by: Bill Erickson <berickxx@gmail.com>

LP#1369169: Add generated README.html to the .gitignore file for OpenSRF

Signed-off-by: Ben Shum <bshum@biblio.org>
Signed-off-by: Bill Erickson <berickxx@gmail.com>

LP#1002028: set Access-Control-Expose-Headers

This allows the OpenSRF JavaScript client library (or
to be precise, one that has been modified to direct
requests at a different domain) to take advantage of CORS
support.

Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1002028: support CORS for HTTPS

Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1002028: Cross Origin Resource Sharing for OpenSRF

Background
----------
Browsers' same-origin policy currently restricts requests to the current
website's domain to prevent various nefarious scenarios. However,
because APIs and other web resources need to remain open to cross-site
use Cross Origin Resource Sharing (CORS) was created to allow services
to formally authorize cross-origin requests. CORS makes it simple to use
OpenSRF's HTTP translator and gateway APIs on websites using separate
domains.

Example Scenarios
-----------------
1) A library would like an AJAX-driven "quicksearch" box on their main
site, which is hosted on a different domain than their catalog.
2) A developer wants to create new web applications and services that
tie into Evergreen, but does not wish to install EG locally or
configure a proxy.

Implementation
--------------
The function crossOriginHeaders() has been added to apachetools.c.
Incoming requests are checked to see if they have an Origin header. The
value of the Origin header is checked against a whitelist defined in
opensrf_core.xml config (XPath: /config/gateway/cross_origin/origin).
The function returns 1 if CORS headers have been added to the response.

Notes
-----
* The OpenSRF Javascript client library (opensrf.js) defaults to the root
of the current web host "/osrf-http-translator." In addition, synchronous
requests are presumed in some situations: resulting in the oncomplete
method never returning (Blocking requests are not possible with cross-
domain XHR.)
* It is also possible to enable CORS with the Apache "set header"
configuration directive. However, this means that the necessary headers
would be appended to every response.

Links
-----
Specification - http://www.w3.org/TR/cors/
Wikipedia Article - http://en.wikipedia.org/wiki/Cross-origin_resource_sharing

Signed-off-by: Bennett Goble <nivardus@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

Revert "LP#1350457: Pass caller's session to subrequests called via method_lookup"

This reverts commit 96917538c6b6fad688dd40c2a8f0cc9127053b24.

(doc) Reorder changes to ejabberd.cfg in install instructions

I found it annoying that the list of changes to make to ejabberd.cfg
didn't follow the order that the options showed up in the default
Debian ejabberd.cfg.  I reordered them so after you finish changing one
option, you can search forward in the document for the next term.

(From Github pull request #3.)

Signed-off-by: Josh Stompro <Stomproj@larl.org>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1350457: Pass caller's session to subrequests called via method_lookup

In the process of looking up a method for an internal subrequest, we lose
session info. This is a problem when the subrequest makes a remote request,
because then the subrequest can't look up the proper locale, among other
things. The forthcoming branch passes the caller's session to the subrequest.

Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1341687 listeners log/drop XMPP error msgs

The most common form of XMPP error messages are "bounced" messages, i.e.
those where the recipient is not available.  Instead of passing these
useless and confusing messages down to a drone for processing, log the
error in the listener and drop the message.

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1343578: document <logtag> in the example opensrf_core.xml

Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1343578: Perl/C syslog "logtag" additions.

 * Honor logtag value when applying a default appname for Perl and C
 * Extract / apply logtag in the router
 * Minor repair: avoid redundant strdup

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1343578: Add support for log tagging

The ability to provide instance-specific log tagging via syslog
would make running multiple instances much easier to manage.  So
this branch does that.

Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1316245: JS now uses browser-native JSON routines

* Replace JSON2jsRaw() with JSON.parse()

* Replace js2JSONRaw() with JSON.stringify()

* Removed unit tests for JSON2jsRaw() and js2JSONRaw()

* Removed jsonPretty() pretty-printing routine.  Use
  JSON.stringify(json, null, <spaces>) instead.

See also:

https://developer.mozilla.org/en-US/docs/Web/JavaScript/Guide/Using_native_JSON

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: disable shared workers pending browser issues

There appears to be a bug in Chromium where loading the same page
multiple times (without a refresh or cache clear) causes the
SharedWorker to fail to instantiate on every other page load.
Further research pending.  Disabling SharedWorker's entirely for
now.

Note, to replicate, load a page using shared workers, focus the
browser address bar, hit Enter to load the page again.  The shared
worker will fail to load on every other page load, though it will
appear to the SharedWorker caller (opensrf.js) that the port is open.

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: update JS/WS/SSL code comment

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: JS status codes can come across as numbers; stringify for match

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: JS libs capture all method errors

Instead of selecting specific errors to report as method errors, report
all API calls which return 400 or 500-series errors as method errors.

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket: avoid sharedworker for firefox 29

https://bugzilla.mozilla.org/show_bug.cgi?id=504553#c73

Avoid using SharedWorkers when useragent match "Firefox".  FF supports
shared workers, but it does not yet support WebSockets within shared
workers.

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets: detect connectedness of JS default sockets

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets: auto-upgrade to shared workers; SSL-always

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets: gateway code repairs & confing options

* avoid unneccessary and wrong incantation of apr_thread_exit.  The two
  sub-threads now both live for the duration of the process.
* to be safe, create thread mutex before threads

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets: apply syslog name in gateway

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets: README typo repairs

Patch from Warren Layton to repair path to 'envvars' and clarify some
text.

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets : additional apache config docs

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets: apache conf -> info logging

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets: remove single-tab JS WS implementation

It was falling behind the shared lib in bug fixes and features.  A
per-tab WS implementation is (maybe) a dangerous thing to have around,
as well, since it encourages /many/ connections.  Can resurrect later if
needed.

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets: shared worker path; JS api_level

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets: initial C libs api_level support

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets : gateway log inbound messages at internal

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket : add JS lib to makefile

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket: avoid module auto configuration

We don't want osrf_websocket_translator to be directly loaded as a
module, since it is not an apache module, but a shared library loaded by
an apache module (mod_websockets).  This is especially true of the default
apache instance.

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket : apache gateway minor fixes

Replace remaining remote_ip calls with a get_client_ip() function which
operates for apache2 and apache4.

Make log functions for applying a log trace value accept const char*'s

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket : more JS api docs

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets : added config docs to install readme

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket JS libs: reconnect and auto-connect

JS clients no longer need to explicitly call the shared WS setup
routine.  It happens under the covers now.

JS client detects disconnects and reconnects when the next message send
attempt occurs.

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket translator idle timeout / graceful shutdown

Added support for an idle timeout and idle check interval configuration
variables.  These allow each websocket apache process to kick off
clients that have been connected and are idle for too long, thus hogging
a process unnecessarily.

Added a SIGUSR1 signal handler which forces the idle timeout to be very
low and a short re-check period so that the client can be kicked as soon
as there are no open conversations.

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets: support WS via shared web workers

This allows mutiple browser tabs to share the same websocket connection.

TODO: add tab (port) disconnect handler
TODO: more docs / examples

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket: do not disconnect ws() on osrf disconnect message

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: temporary websocket installer README

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: Sample Websocket translator Apache 2.2 configuration

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket gateway: improved memory mgt; logging

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websockets; free temporary osrf msgs; minor comment change

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket; docs, more memory mgmt

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket translator

* starting packet inspection
* activity log; recipient removal
* only cache connected recipients; use request_rec pool for session_pool parent
* wrap all thread work in mutex
* session memory goodness

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket JS additions

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: track websocket_plugin.h locally until replaced by makefile.install process

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: websocket gateway repairs and cleanup

* use jsonObjectFree() on jsonObjets, not free();
* removed some debugging logs
* accommodate API changes for Apache 2.4
* safer logging:

  Avoid using ap_log_rerror, in particular referencing server->request
  from the responder thread, since the request_rec will be invalid after
  on_disconnect is called.

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: Apache websocket translator module

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1268619: OpenSRF JS websockets plugin

Adds an opensrf_ws.js handler.  Requries some small modifications to
opensrf.js.  Load opensrf_ws.js in DojoSRF.

For now, to use, add to script:

OpenSRF.Session.transport = OSRF_TRANSPORT_TYPE_WS;

Signed-off-by: Bill Erickson <berick@esilibrary.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1337401: Only care about our own processes

OpenSRF can run mutiple times, as different users, on one host.
Right now we look for all service processes, but we should only
look for our own. This patch does that.

Signed-off-by: Mike Rylander <mrylander@gmail.com>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1315525: Add trusty target and remove lucid from README

Signed-off-by: Ben Shum <bshum@biblio.org>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1315525: Updating Makefile.install for Ubuntu 14.04. Removing lucid support.

Signed-off-by: Chris Sharp <csharp@georgialibraries.org>
Signed-off-by: Ben Shum <bshum@biblio.org>
Signed-off-by: Galen Charlton <gmc@esilibrary.com>

LP#1286248: remove osrf_ctl.sh

As osrf_ctl.sh was deprecated in favor of osrf_control, this patch
removes the deprecated script entirely.

Signed-off-by: Galen Charlton <gmc@esilibrary.com>
Signed-off-by: Bill Erickson <berick@esilibrary.com>

LP#1285915: document that perl2JSON doesn't order hash keys

This patch documents that JSON strings returned by
OpenSRF::Utils::JSON->perl2JSON() should not be expected to have
hash keys sorted in any particular order.  It also adjusts a
corresponding test case to consistently pass under Perl 5.18, which
introduces hash order randomization.

Forcing JSON output to be in canonical form is another option, and
easily done by JSON::XS, but would add overhead.

Signed-off-by: Galen Charlton <gmc@esilibrary.com>
Signed-off-by: Bill Erickson <berick@esilibrary.com>

LP#1234816: improve const-correctness of osrfCachePutString and osrfCachePutObject

Since the cache key is not modified by osrfCachePutString and
osrfCachePutObject, this patch changes the key parameter of those
two functions from char* to const char*.  It also updates one
caller osrfCachePutObject to not cast away const-ness.

This patch has no functional impact, but enables future callers
of osrfCachePut* to pass constant strings without having to
cast away the const-ness.

Signed-off-by: Galen Charlton <gmc@esilibrary.com>
Signed-off-by: Bill Erickson <berick@esilibrary.com>