Re: the Athena trick vs. the Sound trick
On Thu, 23 Mar 2000 02:16:31 GMT, "Stephen J. Turnbull"
<turnbull(a)sk.tsukuba.ac.jp> said:
>>>>> "Hrvoje" == Hrvoje Niksic
<hniksic(a)iskon.hr> writes:
>> * Each time a new inclusion macro needs to be written, you have
>> to redo the AIX test. I concede that this is not very
>> satisfactory.
Hrvoje> That kills it IMO. AIX /per se/ is of course not the
Hrvoje> issue. The issue is that only God knows how many more
Hrvoje> compilers have exactly the same problem. These #ifdef's
Hrvoje> scattered through code might only increase.
But at least the build will fail in an obvious way. That won't be
true if you fail to fix all the Makefile.in.in special rules, as I
understand it. There, you're likely to run into an already fixed
compiler bug, right?
Oh, I *wish* it had failed in an obvious way. ;)
What it *did* was complain about illegal binary characters in
the source code. Seems the IBM compiler *first* botched the
catenation of tokens, and decided that < X11 / foobar.h > could
be parsed as just 'X11'.
Then it said "Hey, I can open() a file called X11 in /usr/include".
So it read it in. Didn't bother doing an fsck'in stat() first.
Ya know what? The bit patterns for the inode numbers do NOT constitute
legal characters in a C program.
Yes, they were having a 2-for-1 sale on bugs that day. ;)
/Valdis