325728 - changes to markerWidth/markerHeight and orient not handled properly

Status: RESOLVED FIXED

(Core :: SVG, defect)

Description

[Robert Longson [:longsonr]]

See bug 320622 comment 10...

> And for that matter we should be checking for changes to markerWidth/Height in
> DidModifySVGObservable and calling SetCoordCtxRect there too. (And while on the
> subject of DidModifySVGObservable it should only be doing the sync mOrientAngle
> stuff when 'observable' is mOrientType. Note we aren't currently observing
> mMarkerWidth/Height, so you'd need to NS_ADD_SVGVALUE_OBSERVER() them in
> Init().)

Also orient cannot be changed to an angle. Since it is mapped to an enum,
setting to an angle it is rejected by the enum code and the change does not
take place.

Comment 1

[Robert Longson [:longsonr]]

Attachment: track changes to markerWidth/markerHeight and fix orient

Comment 2

[Robert Longson [:longsonr]]

Attachment: testcase demonstrating orient issue - click on line

Comment 3

[Robert Longson [:longsonr]]

Attachment: attach working version

Comment 4

[Robert Longson [:longsonr]]

Attachment: add missing optimisation

Comment 5

[Robert Longson [:longsonr]]

Comment on attachment 210612 [details] [diff] [review]
add missing optimisation

optimisation breaks things; currently investigating.

Comment 6

[Robert Longson [:longsonr]]

Attachment: make it actually work with optimisation

GetAttr does not work while the document is loading. Fortunately its possible to get the value directly. Also checked that nsSVGAngle and nsSVGOrient call WillModifySVGObservable and DidModifySVGObservable correctly.

Comment 7

[Robert Longson [:longsonr]]

Comment on attachment 211007 [details] [diff] [review]
make it actually work with optimisation

need to wait for bug 327437 to be checked in now

Comment 8

[Robert Longson [:longsonr]]

Attachment: update for bug 327437 and with memory reduction

Comment 9

[tor]

Comment on attachment 212734 [details] [diff] [review]
update for bug 327437 and with memory reduction

Nice.

> nsSVGMarkerElement::~nsSVGMarkerElement()
> {
>+  if (mMarkerWidth) {
>+    NS_ADD_SVGVALUE_OBSERVER(mMarkerWidth);
>+  }
>+  if (mMarkerHeight) {
>+    NS_ADD_SVGVALUE_OBSERVER(mMarkerHeight);
>+  }

You probably meant NS_REMOVE... in those two cases.

> NS_IMETHODIMP
> nsSVGMarkerElement::DidModifySVGObservable (nsISVGValue* observable,
>                                             nsISVGValue::modificationType aModType)
> {
> #ifdef DEBUG
>-  printf("markerelement - viewport/viewbox/preserveAspectRatio have been changed\n");
>+  printf("markerelement - markerWidth/markerHeight/viewport/viewbox/preserveAspectRatio have been changed\n");
> #endif
> 
>   mViewBoxToViewportTransform = nsnull;
> 
>-  // need to sync mOrientAngle
>-  nsAutoString value;
>-  nsresult rv = GetAttribute(NS_LITERAL_STRING("orient"), value);
>-  if (NS_SUCCEEDED(rv)) {
>-    nsCOMPtr<nsISVGValue> target = do_QueryInterface(mOrientAngle);
>-    target->SetValueString(value);
>-  }
>+  // sync coordinate context with viewbox:
>+  nsIDOMSVGRect *viewbox;
>+  mViewBox->GetBaseVal(&viewbox);
>+  SetCoordCtxRect(viewbox);

Might want to add a test for observable being the viewBox, markerWidth, or MarkerHeight before doing the sync.

Comment 10

[Robert Longson [:longsonr]]

Attachment: address review comments

Comment 11

[tor]

Comment on attachment 212869 [details] [diff] [review]
address review comments

> nsSVGMarkerElement::~nsSVGMarkerElement()
> {
>+  if (mMarkerWidth) {
>+    NS_REMOVE_SVGVALUE_OBSERVER(mMarkerWidth);
>+  }
>+  if (mMarkerHeight) {
>+    NS_REMOVE_SVGVALUE_OBSERVER(mMarkerHeight);
>+  }
>   if (mPreserveAspectRatio) {
>     NS_REMOVE_SVGVALUE_OBSERVER(mPreserveAspectRatio);
>   }
>   if (mViewBox) {
>     NS_REMOVE_SVGVALUE_OBSERVER(mViewBox);
>   }
>-  if (mOrientType) {
>-    NS_REMOVE_SVGVALUE_OBSERVER(mOrientType);
>-  }
> }
> 
> nsresult
> nsSVGMarkerElement::Init()
> {
...
>   // add observers -------------------------- :
>+  NS_ADD_SVGVALUE_OBSERVER(mMarkerWidth);
>+  NS_ADD_SVGVALUE_OBSERVER(mMarkerHeight);
>   NS_ADD_SVGVALUE_OBSERVER(mViewBox);
>   NS_ADD_SVGVALUE_OBSERVER(mPreserveAspectRatio);
>-  NS_ADD_SVGVALUE_OBSERVER(mOrientType);
> 
>   return NS_OK;
> }

I forgot that mapped values are implicitly set as observers, so they don't need to be manually added and removed.

Comment 12

[Robert Longson [:longsonr]]

Attachment: address additional review comments and remove misleading debug output

Comment 13

[tor]

Comment on attachment 212920 [details] [diff] [review]
address additional review comments and remove misleading debug output

> nsSVGMarkerElement::~nsSVGMarkerElement()
> {
>-  if (mPreserveAspectRatio) {
>-    NS_REMOVE_SVGVALUE_OBSERVER(mPreserveAspectRatio);
>-  }
>-  if (mViewBox) {
>-    NS_REMOVE_SVGVALUE_OBSERVER(mViewBox);
>-  }
>-  if (mOrientType) {
>-    NS_REMOVE_SVGVALUE_OBSERVER(mOrientType);
>-  }
> }

If we're going to remove all the guts of the destructor, might as well remove the  method entirely.  With that change, r=tor.

Comment 14

[Robert Longson [:longsonr]]

content has empty destructors in lots of files:
nsSVGCircleElement.cpp
nsSVGDefsElement.cpp
nsSVGTSpanElement.cpp
etc, etc.

I assume these could/should be fixed throughout but I think that's a separate issue. Can we go with this patch as-is and have a followup bug for the removal of empty destructors from non-inherited classes throughout content.

Comment 15

[tor]

Ok, please file a bug about removing those stub methods.

Comment 16

[Robert Longson [:longsonr]]

bug 328433 raised for the stub destructor issue.

Comment 17

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Comment on attachment 212920 [details] [diff] [review]
address additional review comments and remove misleading debug output

+  if (aValueAsString.Equals(NS_LITERAL_STRING("auto")))
...
+  if (value.Equals(NS_LITERAL_STRING("auto"))) {
...
+  if (!value.Equals(NS_LITERAL_STRING("auto")))

use EqualsLiteral

Comment 18

[Robert Longson [:longsonr]]

Attachment: address superreview comments

Comment 19

[tor]

Checked in for Robert.