| All Forums |
Register |
Login |
Search |
Subscriptions |
My Profile |
Inbox |
| Tool Warehouse |
FAQs |
Resources |
Help |
Member List |
Address Book |
Logout |
|
|
gcc 3.4.3 builds but fail a lot of tests (because of what? PIC?)
|
Logged in as: Guest |
| Users viewing this topic: none |
|
Login  |
|
|
 gcc 3.4.3 builds but fail a lot of tests (because of wh... - Dec. 30, '04, 1:26:48 PM
|
|
|
cgrozea
Posts: 4
Joined: Dec. 30, '04,
Status: offline
|
I have tried to compile octave.
I run into troubles with "ambiguous operator overloading".
Searching the net I found that this is some bug in gcc 3.3 fixed only in 3.4.
So, I have built gcc 3.4.3 (latest stable).
It builds almost ok (at some point you have to go into a libtool and replace #! sh with #!/bin/sh).
I tried then a "make check".
This produces tens/hundreds of unexpected FAILs (starting with conversions, then numerics/ieee, then io, etc.).
I'll quote at the end a very small part (what I have now on screen).
Questions:
1. Have anyone else tried to compile (with the GCC 3.3 on SFU) a newer GCC?
2. Is there any way to fix it? Using MSVC at some stage, maybe?
3. I see people say here "Microsoft provided me with a new ld...". Shouldn't one be able to build a new gcc or a new ld using gcc and/or MSVC?
----
sample
FAIL: 27_io/basic_filebuf/sungetc/char/1-in.cc execution test
FAIL: 27_io/basic_filebuf/sungetc/char/1-io.cc execution test
FAIL: 27_io/basic_filebuf/sungetc/char/1-out.cc execution test
FAIL: 27_io/basic_filebuf/sungetc/char/2-in.cc execution test
FAIL: 27_io/basic_filebuf/sungetc/char/2-io.cc execution test
FAIL: 27_io/basic_filebuf/sungetc/char/2-out.cc execution test
FAIL: 27_io/basic_filebuf/sync/char/1.cc execution test
FAIL: 27_io/basic_filebuf/sync/char/1057.cc execution test
FAIL: 27_io/basic_filebuf/sync/char/9182-1.cc execution test
FAIL: 27_io/basic_filebuf/underflow/10096.cc execution test
FAIL: 27_io/basic_filebuf/underflow/char/1.cc execution test
FAIL: 27_io/basic_filebuf/underflow/char/10097.cc execution test
FAIL: 27_io/basic_filebuf/underflow/char/2.cc execution test
FAIL: 27_io/basic_filebuf/underflow/char/9027.cc execution test
FAIL: 27_io/basic_fstream/4.cc execution test
FAIL: 27_io/basic_fstream/rdbuf/char/2832.cc execution test
FAIL: 27_io/basic_ifstream/4.cc execution test
FAIL: 27_io/basic_ifstream/cons/char/1.cc execution test
FAIL: 27_io/basic_ifstream/open/char/1.cc execution test
FAIL: 27_io/basic_ifstream/rdbuf/char/2832.cc execution test
FAIL: 27_io/basic_ios/3.cc execution test
FAIL: 27_io/basic_ios/4.cc execution test
FAIL: 27_io/basic_ios/clear/char/1.cc execution test
FAIL: 27_io/basic_ios/cons/char/1.cc execution test
|
|
|
|
|
RE: gcc 3.4.3 builds but fail a lot of tests (because o... - Dec. 30, '04, 1:32:34 PM
|
|
|
cgrozea
Posts: 4
Joined: Dec. 30, '04,
Status: offline
|
I said hundreds - maybe there are tousends of failures...
=== libstdc++ Summary ===
# of expected passes 1135
# of unexpected failures 883
# of expected failures 4
# of unsupported tests 8
|
|
|
|
|
RE: gcc 3.4.3 builds but fail a lot of tests (because o... - Dec. 30, '04, 1:52:23 PM
|
|
|
Rodney
Posts: 2949
Joined: Jul. 9, '02,
From: /Tools lab
Status: offline
|
You need to look at the changes that were done to 3.3 and bring them forward
to the gcc you are using. The diff's are on the CD.
The changes to gcc were submitted a few years ago and a particular company well
versed with gcc was supposed to see this happen. But it didn't.
PIC/pic support was never done for Interix (and it doesn't need it anyway). But specifying
PIC/pic can at times produce bad code. So PIC/pic shouldn't be used.
|
|
|
|
|
RE: gcc 3.4.3 builds but fail a lot of tests (because o... - Jan. 2, '05, 5:20:57 AM
|
|
|
cgrozea
Posts: 4
Joined: Dec. 30, '04,
Status: offline
|
Thank you for answer.
Not sure I understand:
- what CD?
- see what happen? patches applied?
Some more attempts I did:
1. Tried to set CC=cc (visual 7.1) and then make bootstrap gcc.
This fails buildling libiberty, not accepting 0/0 as NAN but producing a compile time error instead.
2. Tried to compile Octave with cc, also fails at some piece of code that looks valid at the first sight (and which compiled ok with gcc).
3. Thinking of trying to build gcc 3.4.3 from within interix using an external win32 compiler which is also a GCC (maybe mingw).
I think I could try to immitate the way cc calls the visual c++ compiler.
Should this work or there is something fundamental what I don't understand about Interix/SFU?
Cygwin gets all unix-like environment from a DLL, how are those provided to interix programs? Libraries? Or something different - like just running under the reworked posix environment?
|
|
|
|
|
RE: gcc 3.4.3 builds but fail a lot of tests (because o... - Jan. 2, '05, 6:53:16 PM
|
|
|
Rodney
Posts: 2949
Joined: Jul. 9, '02,
From: /Tools lab
Status: offline
|
> Not sure I understand:
> - what CD?
> - see what happen? patches applied?
When you installed Interix/SFU you used a CD or you downloaded a CD image. Either
on the CD or the extracted CD image you will find the sources/diff's for GNU
applications. These diff's are the changes applied to gcc (and associated code)
for gcc/g++/g77 to work on Interix.
Included with the gcc diffs are some notes that you should read. this
should help with understanding what's been done and to what code. It should
also help with getting a built going.
> Cygwin gets all unix-like environment from a DLL, how are those provided to interix programs?
> Libraries? Or something different - like just running under the reworked posix environment?
Cygwin is an emulation environment for the system calls (that's what the cygwin DLL does:
converts Unix calls to Win32-speak and back again).
Interix is a subsystem over the NT kernel (same as Win32; they are peers). So no translation
to Win32-speak needed. Directly talking to the kernel means a number of things are way faster
(such as fork()).
|
|
|
|
|
RE: gcc 3.4.3 builds but fail a lot of tests (because o... - Jan. 18, '05, 2:23:11 PM
|
|
|
jumphigh
Posts: 61
Joined: Jul. 10, '04,
From: Germany
Status: offline
|
Hi Rodney!
I only loaded the SFU 3.5 download. But that is not a CD image. It extracts in a file system. Are these diffs included inside?
What's the problem about building a new gcc? I believe you only have to change the linker to build PE executables with POSIX as subsystem. Or am I wrong? Is it impossible to use e new gcc with the old libs provided by SFU. What's about a new C++ standard lib and STL?
Regards,
Andreas
|
|
|
|
|
RE: gcc 3.4.3 builds but fail a lot of tests (because o... - Jan. 18, '05, 2:54:52 PM
|
|
|
Rodney
Posts: 2949
Joined: Jul. 9, '02,
From: /Tools lab
Status: offline
|
> I only loaded the SFU 3.5 download. But that is not a CD image. It extracts in a file system. Are these diffs included inside?
Yes they are. Everything on the CD is in that file extraction.
> I believe you only have to change the linker to build PE executables with POSIX as subsystem. Or am I wrong?
There's more to it than that. There are changes in utilities and libraries that build gcc.
All of the changes are exposed in the diff's.
> Is it impossible to use e new gcc with the old libs provided by SFU. What's about a new C++ standard lib and STL?
I don't know. There was someone a while back that who did a build of just gcc using the older libraries
and utilities. It was posted here in the forum waht they did. You can search for it. Set the time frame to
within the last year.
|
|
|
|
 New Messages |
 No New Messages |
 Hot Topic w/ New Messages |
 Hot Topic w/o New Messages |
|
 Locked w/ New Messages |
 Locked w/o New Messages |
|
 Post New Thread
Reply to Message
Post New Poll
Submit Vote
Delete My Own Post
Delete My Own Thread
Rate Posts |
|
|
|
Printable Version
