|
View:
New views
2 Messages
—
Rating Filter:
Alert me
|
|
|
[gzip-1.3.13] getopt tool chain build problem and other new problemsI've successfully installed gzip-1.3.13 today on nearly 20 flavors of
Unix. However, I'm getting failures on some platforms that I also saw today in a test release of the zile editor that look like an error in the getopt component of the GNU tool chain. Specially, on Mac OS X (PowerPC and Intel), FreeBSD 5.0 and 6.1, NetBSD 1.6, I get compilation failure like this: CC getopt.o In file included from ./getopt.h:28, from getopt.h:28, from getopt.c:26: /usr/include/getopt.h:53: error: redefinition of `struct rpl_option' /usr/include/getopt.h:68: error: conflicting types for 'rpl_getopt_long' ./getopt.h:227: error: previous declaration of 'rpl_getopt_long' was here /usr/include/getopt.h:68: error: conflicting types for 'rpl_getopt_long' ./getopt.h:227: error: previous declaration of 'rpl_getopt_long' was here I can reproduce this with gcc-2.95.3, gcc-3.4.3, gcc-4.1, gcc-4.2, gcc-4.3, gcc-4.4, and gcc-4.5, so it does not appear to be compiler dependent. On Solaris 7 SPARC, I get this failure with native cc: CC getopt.o In file included from ./getopt.h:28, from getopt.h:28, from getopt.c:26: /usr/include/getopt.h:53: error: redefinition of `struct rpl_option' /usr/include/getopt.h:68: error: conflicting types for 'rpl_getopt_long' ./getopt.h:227: error: previous declaration of 'rpl_getopt_long' was here /usr/include/getopt.h:68: error: conflicting types for 'rpl_getopt_long' ./getopt.h:227: error: previous declaration of 'rpl_getopt_long' was here Switching to gcc-2.95.3 produces a different failure: CC lstat.o In file included from lstat.c:43: sys/stat.h:343: warning: `struct stat' declared inside parameter list sys/stat.h:343: warning: its scope is only this definition or declaration, which is probably not what you want. lstat.c:62: warning: `struct stat' declared inside parameter list lstat.c:63: conflicting types for `rpl_lstat' sys/stat.h:326: previous declaration of `rpl_lstat' lstat.c: In function `rpl_lstat': lstat.c:65: warning: passing arg 2 of `orig_lstat' from incompatible pointer type lstat.c:76: dereferencing pointer to incomplete type lstat.c:84: dereferencing pointer to incomplete type lstat.c:89: warning: passing arg 2 of `rpl_stat' from incompatible pointer type I have gzip-1.3.12 installed on all of these systems, so the problem is new with 1.3.13. Finally, please eliminate the cryptic compilation reports like this: CC frexpl.o CC getopt.o CC getopt1.o CC isnanf.o This is distinctly unhelpful, and covers up what is really happening. make's normal output is what should be appearing, with the complete compiler command. ------------------------------------------------------------------------------- - Nelson H. F. Beebe Tel: +1 801 581 5254 - - University of Utah FAX: +1 801 581 4148 - - Department of Mathematics, 110 LCB Internet e-mail: beebe@... - - 155 S 1400 E RM 233 beebe@... beebe@... - - Salt Lake City, UT 84112-0090, USA URL: http://www.math.utah.edu/~beebe/ - ------------------------------------------------------------------------------- |
|
|
Re: [gzip-1.3.13] getopt tool chain build problem and other new problemsNelson H. F. Beebe wrote:
> I've successfully installed gzip-1.3.13 today on nearly 20 flavors of > Unix. > > However, I'm getting failures on some platforms that I also saw today > in a test release of the zile editor that look like an error in the > getopt component of the GNU tool chain. > Specially, on Mac OS X (PowerPC and Intel), FreeBSD 5.0 and 6.1, > NetBSD 1.6, I get compilation failure like this: > > CC getopt.o > In file included from ./getopt.h:28, > from getopt.h:28, > from getopt.c:26: > /usr/include/getopt.h:53: error: redefinition of `struct rpl_option' > /usr/include/getopt.h:68: error: conflicting types for 'rpl_getopt_long' > ./getopt.h:227: error: previous declaration of 'rpl_getopt_long' was here > /usr/include/getopt.h:68: error: conflicting types for 'rpl_getopt_long' > ./getopt.h:227: error: previous declaration of 'rpl_getopt_long' was here Thanks for the report. I think that's just been fixed via gnulib, so I may make a new release soon. ... > Finally, please eliminate the cryptic compilation reports like this: > > CC frexpl.o > CC getopt.o > CC getopt1.o > CC isnanf.o > > This is distinctly unhelpful, and covers up what is really happening. > make's normal output is what should be appearing, with the complete > compiler command. That's just the default. You can get the old behavior via ./configure --enable-silent-rules ... && make or ./configure ... && make V=1 |
| Free embeddable forum powered by Nabble | Forum Help |