eric@CSE.OGI.EDU (Eric Hanchrow) (01/11/90)
gcc version 1.35 Sun 386i SunOS 4.0 This isn't the most critical bug in the world, but it's sort of annoying. I ran the `fixincludes' script when I installed gcc, and everything seemed fine from then on... And when I used gcc to compile Sunview applications, and saw errors like `/usr/include/sundev/vuid_event.h: Garbage at end of ifdef', I figured that meant I ought to compile with the `-traditional' flag. I did, and then all was well. But then I figured, why didn't `fixincludes' fix that include? I looked at the directory `/usr/local/lib/gcc-include', and saw that the sundev directory was a symbolic link, namely `sundev -> ../cluster/devel/base_devel/include/sundev' Now since that's a relative pathname, the link really doesn't point to anything real. So I suppose when gcc is searching for vuid_event.h, it looks there first, finds that there is no such directory as /usr/local/lib/cluster ..., and just keeps searching, until it finds the original Sun-supplied file in /usr/include. Then since that include file is non-ANSI, it complains. I don't have the script around to look at (hadda save disk space and put source onto tapes), but perhaps it uses `cp -r' to build the gcc-include directory. This is usually a disaster on Suns, what with their forests of symbolic links. `tar' would probably work better. So I held my breath and copied the gcc-include directory from our Sun4 onto the Sun386. Hope that doesn't cause obscure bizarre problems. Anyhow, you might want to re-think that script in future releases. ------------------------------------------------------------------------- |Eric Hanchrow yamada-sun!eric@nosun.west.sun.com | |Phase III Logic, Inc. ...!{tektronix, sun}!nosun!yamada-sun!eric | |1600 N.W. 167th Place Beaverton, OR 97006 | |Voice: (503)-645-0313 Fax: (503)-645-0207 as of 13-Dec-89 | -------------------------------------------------------------------------