alexis@panix.uucp (Alexis Rosen) (06/07/91)
When installing the latest gcc binaries from wuarchive, I ran the
fixincludes script, which produced (among many other messages) the following:
Fixing sys/file.h, bad
"/usr/local/Gnu/lib/gcc-include/sys/file.h" 150 lines, 4789 characters
Substitute pattern match failed
Deleting /usr/local/Gnu/lib/gcc-include/sys/file.h; no fixes were needed.
Any ideas what this means? I took a look at file.h and didn't see anything
obviously wrong. (Although that doesn't mean much...)
Thanks,
---
Alexis Rosen
Owner/Sysadmin, PANIX Public Access Unix, NY
alexis@panix.com
{cmcl2,apple}!panix!alexisrmtodd@servalan.uucp (Richard Todd) (06/09/91)
alexis@panix.uucp (Alexis Rosen) writes: >When installing the latest gcc binaries from wuarchive, I ran the >fixincludes script, which produced (among many other messages) the following: >Fixing sys/file.h, bad >"/usr/local/Gnu/lib/gcc-include/sys/file.h" 150 lines, 4789 characters >Substitute pattern match failed >Deleting /usr/local/Gnu/lib/gcc-include/sys/file.h; no fixes were needed. >Any ideas what this means? I took a look at file.h and didn't see anything >obviously wrong. (Although that doesn't mean much...) What this means is that they changed sys/file.h between A/UX 2.0 and A/UX 2.0.1, and the changed version is confusing fixincludes--fixincludes is being made to think that there's something in there that might need fixing, but there isn't. You'll probably have to look thru sys/file.h and see if there's anything there that's obviously ANSI-incompatible and fix it by hand. Most of what fixincludes is trying to fix is the old-style "Reiserisms", i.e. preprocessor tricks to do token pasting and the like that worked with the old Reiser CPP, but must be done differently under ANSI C. Example (from sys/ioctl.h, the home of most of the nasty Reiserisms): ------- #define _IO(x,y) (IOC_VOID|('x'<<8)|y) ... #define TIOCSETRDSIG _IO(t, 106) /* pty: set read signal */ ------- Note that the argument to the _IO macro appears inside the '' quotes, so the _IO(t,106) is supposed to (under old CPP) expand to (IOC_VOID|('t'<<8)|106). Sticking macro arguments inside '' like this is verboten under ANSI C; the ANSI equivalent looks like this: ------- #define _IO(x,y) (IOC_VOID|(x<<8)|y) ... #define TIOCSETRDSIG _IO('t', 106) /* pty: set read signal */ ------- See the difference? Now just look for things like that in sys/file.h and see what you find. Frankly, looking at the A/UX 2.0 sys/file.h, I don't see anything in there that needs fixing; I think fixincludes is just confused. (Most all that fixincludes does is fix Reiserisms; the rest of fixincludes seems to be fixing some Sun-specific botches in include files that don't affect the rest of us....) -- Richard Todd rmtodd@uokmax.ecn.uoknor.edu rmtodd@chinet.chi.il.us rmtodd@servalan.uucp "We're thinking about upgrading from SunOS 4.1.1 to SunOS 3.5."--Henry Spencer