recent regression

[Bryce McKinlay]

Ranjit Mathew wrote:

>>Would it not it be better to simply avoid calling read_class in the
>>CLASS_FROM_SOURCE_P case? If you have a type decl for a class and
>>CLASS_FROM_SOURCE_P is true, then (afaik) it must by definition be
>>already loaded and been/being parsed. How about the following patch instead:
>>    
>>
>[...]
>  
>
>>This would also allow us to get rid of many of the CLASS_FROM_SOURCE_P
>>checks scattered around parse.y.
>>    
>>
>
>Yes, this is much better. As I mention elsewhere though, we ought
>to really clarify what we mean by a "loaded" class. (i.e. load_class()
>v/s CLASS_LOADED_P().)
>
>Will you be fleshing this out a bit more to remove the redundant
>checks you mention?
>
>Also, if this is the way forward, then you should remove the
>comment at the top of load_class() where the author intends
>to ask clients to call read_class() directly.
>  
>

Agreed. Here's a complete patch with redundant CLASS_FROM_SOURCE_P's 
removed. I've run this through the testsuite (no regressions), and also 
built parts of RHUG against it. It seems to fix PR 16768 as well.

OK to commit?

Bryce


-------------- next part --------------
A non-text attachment was scrubbed...
Name: gcj-PR1207.patch
Type: text/x-patch
Size: 3398 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/java-patches/attachments/20040625/1e9ca4ab/attachment.bin>