| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
A dependency loop exists between class/agent.class.php and
class/group.class.php. Due to specific `require_once` ordering within
the app, this problem was not surfaced until now.
These two classes depend on eachother, but strictly speaking, the
interpreter needs to read the agent class first. This is because group
directly inherits from agent. It is only one of agent's functions which
references group. Group has a `require_once "class/agent.class.php"` at
its top, so requiring group first will read both classes, in the correct
order, and provide their definition's for the remainder of the runtime.
The main entry-point, index.php, did not have this problem since it was
explicitly requiring group itself (it actually needs group, though).
The df.php entry-point wasn't and was relying on requires in the class/
directory to resolve this issue. In a more-sane language, I could patch
this more easily directly in the affected file; rather, this patch
updates the df entry-point to explicitly require group, solving the
issue up front.
Hopefully this can be fleshed out in the future as it should not consern
the entry-points that this specific evaluation order needs to take
place.
The third and final entry-point, cron.php, is already fine at the time
of this commit. Its require tree is much simplier, and does not even
include either of the affected classes.
|
|
|
|
This is probabally more of an oops than a bug, although was causing
unexpected behavior.
When falling back to checking whether the agent has access to the
object's owner, it was wrongly accessing through $this->owner, rather
than $obj->owner (which is the function argument). This was probabally
left over from how this function _used_ to be implemented (you would
call on the object and pass in the user).
|
|
|
|
|
|
|
|
There was a problem with processing enum type fields. The way all other
field types are asserted to be 'defined' is via:
isset($field) && $field != ""
Which works perfectly fine, and is exactly what we want. However, with
enums the second part of that && can bite us if "" is in the list of
acceptable values.
This commit removed that half of the check (only for enum values) so
that the empty string may be an acceptable enum value. If "" is not in
the values array, then the check is implicitly reinstated.
|
|
|
|
Because of how this function was implemented, any failure during
database instance construction is treated the same way. IE. we cannot
tell the difference between 'no db config' (as is the initial default
state) and a 'bad db config' (either bogus data, or the server happens
to be down).
Because of this, if, after the database access is initially set up,
access to the db becomes unavailable or someone makes a bad edit to the
dbconfig.php file, Scrott behaves as if it is being configured for the
first time. This is *dangerous* behavior! (unexpected, at the least)
The implication of this is that if Scrott's database access is ever
incidentially interrupted, the very next visitor to the site is offered
the chance to (silently) reconfigure the server to point to any database
of his choosing.
This patch updates the checkConfig() function to only 'soft fail'
(return false) in the case where the configuration is _actually_
missing. IE. $_SCROTT['conf'] is not defined. This function will
otherwise passthrough any and all exceptions which result from
instanciating the database instance and will only return true if both of
these steps succeed.
|
|
This is an interface to alter the settings stored in the 'settings'
database table. Typically restricted to administrators only.
|
|
This is not meant to be a user (admin) configurable key. Rather, this
is a mechanism for success/failure results from interactions with
PHPMailer to make their way back to the UI. Down the road, email
sending functions should publish their true/false return value to this
configuration key. The initial default value of "NULL" means no mail
send attempts have taken place.
|
|
|
|
|
|
|
|
The controller now (again) prevents browsing to objects the user is not
allowed to access.
|
|
|
|
In cases where the pad had no stages beneath it, `new
stage($this->stage)` would construct an invalid object. As it turns
out, calling ->getArray() on an uninitialized stage object yeilds bad
results.
Instead of patching the stage::getArray() function, I add a check to
harden pad::getStages(). My reasoning for this is as follows: The bug
in getArray() manifests from a domain error, ie. it's only because we
are calling it on an uninitialized object. The object is already in a
bad state prior to caling getArray(). Rather, I opt to patch
getStages() so that we never create a bad object in the first place.
Now, for no-stage pads, getStages() will return early an empty array.
|
|
This particular flaw was dampening (and could popentially be hiding) the
effects of other bugs.
For instance, in this case, a GUID of "" was invalidly being used to
construct an object. This should obviously be considered an error, but
since "" evaluates to false, the construction was treated as default (no
GUID) construction and succedded. It wasn't until later when missing
properties were accessed that random PHP error messages clued me into what
was happening.
Now, when any sort of explicit value is used to construct an object (not
NULL), an object load will be attempted, giving bad input more chances
to fail outright and trigger an exception. In addition, the 'no such
guid' exception message is updated to place quotes ('') around the GUID
string to make it more obvious when "" is used in the future.
|
|
Removed placholder text and now displaying the logged-in user's display
name. This is accopanied by the user icon and (conditionally) the
'cool' sunglasses icon for admins. This is similar, but not identical,
to the markup of my old archived code I'm clearing out.
|
|
This additional spacing is added to the navbar text to improve its
appearance in the browser.
|
|
As a convenience, I'm adding a basic set of stages to new pads. These
stages are "To Do", "In Progress" ('Done' being closed issues). This
implements a very simple workflow for new pads.
|
|
These two functions, 'getCurrentTimestamp()' and 'isGUID()' are updated
to be public. There is actually no good reason for them to be private;
I originally just never antisipated their use outside this class.
I need isGUID() in index.php to help with page routing. Neither of
these two functions have side effects of any kind nor any unexpected
behavior, so there is no harm in going public.
|
|
Removing these unnecessary checks. They are not protecting us from
anything, only inconveniencing me in index.php.
|
|
|
|
Update all usage of saveFile() to use added saveIfFile() function,
forwarding on the convenience to model code. Model code can pass in
file field names, rather than $_FILES arrays directly.
|
|
This is an alternative function to globals' saveFile(), which allows
model code to just pass in the name of the expected uploaded file,
rather than requiring them to look up the file themselves. This is in
line with my preference to encapsulate PHP superglobals access away from
most of the codebase.
Note that even if the user opts not to upload optional files, the
associated file <input> field will still be present in $_FILES, with a
special error code set (meaning 'no file uploaded') which setFile()
ignores. It is only in the case of a malformed form submission that
$_FILES will be missing the requested file field, prompting Scrott to
throw an exception.
|
|
The input[ ] portion of the hidden field name had been left off. This
commit fixes this bug.
|
|
Updated old settings modal design, to incorporate added Scrott features.
Reimplemented this modal in Scrott from archived code and have included
it in the navbar.
|
|
|
|
Similar to objHeadCircle() but for displaying a bg img on page, rather
than as the actual background.
|
|
We can check for the existence of an object's background image by
calling getBgImg(), since it returns NULL when there is no such image.
But getHeadImg() behaves differently, returning a path to
'static/img/null.jpg' (via df.php) when there is no image, making it
more difficult to tell. This function addresses this concern.
|
|
This ctrl is not POSTed as part of the input[] array, but as a
standalone name.
|
|
|
|
|
|
|
|
The initial intended use case for this is applying the "active" and "in
active" classes to the first tab to appear in the settings modal.
|
|
|
|
|
|
This is basically a constructor for agent. The actual type returned is
a contrete agent.
|
|
|
|
|
|
|
|
If a table query yeilds zero rows, we would still attempt to load the
first (index zero) into $this, causing an error to be thrown by PHP. We
are now checking the size of the results array first.
|
|
|
|
|
|
|
|
|
|
UI module to use when listing out groups and pads. These will link to
the group/pad page and display information about the object.
|
|
Added a $limit argument to specify a maximum number of results to
return.
|
|
|
