This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: RFC/RFHelp: c-decl.c rewrite - almost but not quite

Here finally is the big c-decl.c rewrite.  I'm pleased to report it
solves the infamous dwarf2out ICE and several other bugs (see the
changelog for details).  However, I need some help finishing it.

Most significantly, the Objective-C and possibly Java front ends
appear to be broken by this patch, such that neither runtime library
can be built.  For Objective-C the symptoms are a bunch of warnings
about unexpected types for 'id' followed by an ICE in list_length
owing to a list of global decls that has somehow become circular.
(I added circular list detection to list_length under ENABLE_CHECKING;
without that it would be an infinite loop instead of an ICE.)

Considering this is a 4000 line diff, could you give a high-level description of how the code's changed architecturally? Your previous c-decl.c changes did a pretty nice job of explaining what was going on in each change ("another coffee spoon of progress on c-decl.c" from March 9 and "RFC/PATCH: c-decl.c: no getdecls() for you" from Feb. 8, or "c-decl.c overhaul (1/3) Refactor duplicate_decls" from Jan. 10)

We're mucking in potentially similar code on the C++ side for the stree
work, and it would be understand how this work might conflict with


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]