1301249 - nsIDocument::GetDocumentURI() should be fallible

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect, P3)

Description

[Nicholas Nethercote [inactive]]

This is a spin-off of bug 1297300, which is adding checking for GetSpec() calls that are lacking them.

nsIDocument::GetDocumentURI() has a GetSpec() call that is fallible. The right way to handle this is to make nsIDocument::GetDocumentURI() itself fallible. bz noted in bug 1297300 comment 31:

> Make this funcation fallible.  I just looked through the callers and some of 
> them want to throw exceptions to JS if this fails anyway, some shouldn't be
> using this function at all because it causes extra encoding conversions
> (NS_IMPL_CYCLE_COLLECTION_TRAVERSE_BEGIN_INTERNAL(DOMEventTargetHelper),
> restyle logging, style context asserts (via GetURL), global window code
> checking for "about:neterror" (via GetURL) I'm looking at you), some could
> start throwing exceptions to JS even though they don't right now.

Comment 1

[Nicholas Nethercote [inactive]]

And its return value should be MOZ_MUST_USE, of course :)

Comment 2

[Andrew Overholt [:overholt]]

Please let me know if this should be a higher priority than "backlog".

Comment 3

[Nicholas Nethercote [inactive]]

> Please let me know if this should be a higher priority than "backlog".

That seems fine.

Comment 4

[Andrea Marchesini [:baku]]

Attachment: doc1.patch

Comment 5

[Boris Zbarsky [:bzbarsky]]

>+      rv.SuppressException();

Can just use an IgnoredErrorResult.

Comment 6

[Boris Zbarsky [:bzbarsky]]

>+    result.Append(NS_ConvertUTF16toUTF8(url).get());

I know this is preexisting, but...

  CopyUTF16toUTF8(url, result);

Comment 7

[Nicholas Nethercote [inactive]]

Comment on attachment 8791131 [details] [diff] [review]
doc1.patch

Review of attachment 8791131 [details] [diff] [review]:
-----------------------------------------------------------------

I will defer to a DOM peer for r+.

Comment 8

[Olli Pettay [:smaug][[e-mail]]]

Comment on attachment 8791131 [details] [diff] [review]
doc1.patch

I don't see reason for using ErrorResult when GetURL and GetDocumentURI could just return nsresult (possibly with MOZ_MUST_USE).

IMO ErrorResults brings just too complicated setup for this kinds of Gecko internal methods.

Comment 9

[Andrea Marchesini [:baku]]

Attachment: doc1.patch

Comment 10

[Olli Pettay [:smaug][[e-mail]]]

Comment on attachment 8792486 [details] [diff] [review]
doc1.patch

> #define STYLE_STRUCT_INHERITED(name_, checkdata_cb)                            \
>     data = destroyingInheritedData.mStyleStructs[eStyleStruct_##name_];        \
>     if (data &&                                                                \
>         !(aDestroyingContext->mBits & NS_STYLE_INHERIT_BIT(name_)) &&          \
>          (mCachedInheritedData.mStyleStructs[eStyleStruct_##name_] == data)) { \
>       printf_stderr("style struct %p found on style context %p\n", data, this);\
>       nsString url;                                                            \
>-      PresContext()->Document()->GetURL(url);                                  \
>-      printf_stderr("  in %s\n", NS_ConvertUTF16toUTF8(url).get());            \
>-      MOZ_ASSERT(false, "destroying " #name_ " style struct still present "    \
>-                        "in style context tree");                              \
>+      nsresult rv = PresContext()->Document()->GetURL(url);                    \
>+      if (NS_SUCCEEDED(rv)) {                                                  \
>+	printf_stderr("  in %s\n", NS_ConvertUTF16toUTF8(url).get());          \
>+	MOZ_ASSERT(false, "destroying " #name_ " style struct still present "  \
>+			  "in style context tree");                            \
>+      }                                                                        \
align \ characters and I would indent stuff inside the 'if' to keep this stuff readable.
You don't want MOZ_ASSERT to be inside the 'if'.

> #define STYLE_STRUCT_RESET(name_, checkdata_cb)                                \
>         data = destroyingResetData->mStyleStructs[eStyleStruct_##name_];       \
>         if (data &&                                                            \
>             !(aDestroyingContext->mBits & NS_STYLE_INHERIT_BIT(name_)) &&      \
>             (mCachedResetData->mStyleStructs[eStyleStruct_##name_] == data)) { \
>           printf_stderr("style struct %p found on style context %p\n", data,   \
>                         this);                                                 \
>           nsString url;                                                        \
>-          PresContext()->Document()->GetURL(url);                              \
>-          printf_stderr("  in %s\n", NS_ConvertUTF16toUTF8(url).get());        \
>-          MOZ_ASSERT(false, "destroying " #name_ " style struct still present "\
>-                            "in style context tree");                          \
>+          nsresult rv = PresContext()->Document()->GetURL(url);                \
>+          if (NS_SUCCEEDED(rv)) {                                              \
>+            printf_stderr("  in %s\n", NS_ConvertUTF16toUTF8(url).get());      \
>+            MOZ_ASSERT(false, "destroying " #name_ " style struct still present "\
>+                              "in style context tree");                        \
>+          }                                                                    \
MOZ_ASSERT shouldn't be inside the 'if', And align \

Or other option is to explicitly say that return value of GetURL isn't used.
Those fixed, r+

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/aa252c2f85b5
nsIDocument::GetDocumentURI() should be fallible, r=smaug

Comment 12

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/aa252c2f85b5

Comment 13

[Gerry Chang [:gchang]]

Mark 51 as fix-optional. If it's worth uplifting to 51, feel free to nominate it.