Last C++ bootstrap bits
Gabriel Dos Reis
Tue Aug 3 06:57:00 GMT 2004
Bernardo Innocenti <email@example.com> writes:
| Gabriel Dos Reis wrote:
| > Bernardo Innocenti <firstname.lastname@example.org> writes:
| > | Hello,
| > |
| > | I still have 50KB-worth of C++ bootstrap patches, down from
| > | 250KB.
| > |
| > | It's too late to finish splitting them up, applying, testing
| > | and writing ChangeLog entries, so I'll leave it for tomorrow.
| > | I'm posting them now just in case someone wants to comment
| > | on how I addressed the remaining issues.
| > |
| > | These bits address bool-correctness, enum-correctness and
| > | resolve clashes between variables and typenames. I quite
| > | dislike the "rtrtx" and "rtrtvec" member names, so I'm
| > | looking for suggestions for improvement.
| > Hi,
| > My attention was drawn to your patch. Sorry to realize that I
| > essnetially duplicated your work. I think the "bool" issue should be
| > submitted in a different patch, as it does not address compilation
| > with a C++ compiler per se. My last two patches handled the GC enums
| > and rtx issue. I have another patch for the VARRAY issue.
| > I believe we should investigate RTH's suggestion on the
| > enumerationns tree_code and <lang>_tree_code.
| I was hoping to split, test and commit my patch over this
| weekend, but then I've been too busy to do it.
Don't feel bad -- over the last six months, I've been hoping to have
time "next week" :-(
| So I'm glad you finally did these missing bits, even
| though there has been some duplication of effort. I'd
| not have enough time until next weekend anyway.
| I still have these pending issues on my TODO list:
| - the bool conversions (not related with C++, as you say);
| - reverse the "real bool in C++" change, as requested by Zack;
| - add type-safe wrappers like XNEW() for alloca();
| - address the c_tree_code/tree_code problem, as discussed
| with rth.
I've been investigating issues related to
- c_tree_code/tree_code problem, and
- headers (auto-host.h and config.h) generated at configuration
time, that seems to be missing many important information.
For example, rlim_t does not get detected and configure #define to
"long". Later, that #define cause an incomprehensible diagnostic.
Also, some functions from <stdio.h> or glibc additions are nto
correctly detected. Later, their declarations clash with those
provided by the target -- C does not care, but C++ quickly objects
when the accompanying exception specifiction goes wrong.
More information about the Gcc-patches