cnh5730@calvin.tamu.edu (Chuck Herrick) (10/13/90)
Does this appear to be a bug in the C preprocessor to you? With setenv cpp /lib/cpp I ran the following script (extracted from Larry Walls' Configure in patch2) cat <<'EOT' >testcpp.c #define ABC abc #define XYZ xyz ABC.XYZ EOT echo 'Maybe "'$cpp'" will work...' $cpp <testcpp.c >testcpp.out 2>&1 inspection of testcpp.out revealed that the file contained abc .xyz and NOT abc.xyz Since Larry's script checks to see if 'abc.xyz' IS contained, it promptly bombs out. By the way, /lib/cpp is a link to /lib/cpp-16 -- Chuck Herrick cnh5730@calvin.tamu.edu
cnh5730@calvin.tamu.edu (Chuck Herrick) (10/13/90)
In article <9060@helios.TAMU.EDU) cnh5730@calvin.tamu.edu (Chuck Herrick) writes:
)Does this appear to be a bug in the C preprocessor to you?
)I ran the following script (extracted from Larry Walls' Configure
)in patch2)
)cat <<'EOT' )testcpp.c
)#define ABC abc
)#define XYZ xyz
)ABC.XYZ
)EOT
)echo 'Maybe "'$cpp'" will work...'
)$cpp <testcpp.c )testcpp.out 2)&1
)inspection of testcpp.out revealed that the file
)contained
) abc .xyz
)and NOT
) abc.xyz
In addition, I ran the script on a Sun SPARC 1+ which has
version 1.37 of the GNU C compiler/preprocessor
(cccp and cpp) installed along with the
vanilla Sun Berkeley C compiler.
when I ran the Sun C preprocessor (/lib/cpp) I got
abc.xyz
and when I ran GNU's cccp and cpp I got
abc .xyz
Since Larry's patch2 requires a C preprocessor which does
not produce this bug in order to unwrap and configure patch2,
one wonders how one will install a patch onto the GNU
C preprocessor.
--
Chuck Herrick cnh5730@calvin.tamu.edugeoff@ITcorp.com (Geoff Kuenning) (10/15/90)
In article <9063@helios.TAMU.EDU> cnh5730@calvin.tamu.edu (Chuck Herrick) writes: > Since Larry's patch2 requires a C preprocessor which does > not produce this bug in order to unwrap and configure patch2, > one wonders how one will install a patch onto the GNU > C preprocessor. This is an artifact of GCC which Larry knows about. Newer versions of Configure use a slightly different regular expression which will handle the inserted blank space (which is legal under the ANSI spec). Meanwhile, the easy solution is to just modify the regular expression in Configure by hand so that it handles blanks. -- Geoff Kuenning geoff@ITcorp.com uunet!desint!geoff
moose@svc.portal.com (10/16/90)
In article <9060@helios.TAMU.EDU> cnh5730@calvin.tamu.edu (Chuck Herrick) writes: >Does this appear to be a bug in the C preprocessor to you? Yes, but it is NOT a bug. >#define ABC abc >#define XYZ xyz >ABC.XYZ >inspection of testcpp.out revealed that the file >contained > abc .xyz >and NOT > abc.xyz That is correct. That is ANSI not traditional preprocessing. Check out the ## token pasting. -- Michael Rutman | moose@svc.portal.com Cubist | makes me a NeXT programmer Software Ventures | That's in Berkeley smile, you're on standard disclaimer | <fill in with cute saying>
pbiron@weber.ucsd.edu (Paul Biron) (10/16/90)
In article <9063@helios.TAMU.EDU>, cnh5730@calvin.tamu.edu (Chuck Herrick) writes: |> In article <9060@helios.TAMU.EDU) cnh5730@calvin.tamu.edu (Chuck Herrick) writes: |> )Does this appear to be a bug in the C preprocessor to you? |> )I ran the following script (extracted from Larry Walls' Configure |> )in patch2) |> )cat <<'EOT' )testcpp.c |> )#define ABC abc |> )#define XYZ xyz |> )ABC.XYZ |> )EOT |> )echo 'Maybe "'$cpp'" will work...' |> )$cpp <testcpp.c )testcpp.out 2)&1 |> )inspection of testcpp.out revealed that the file |> )contained |> ) abc .xyz |> )and NOT |> ) abc.xyz |> |> In addition, I ran the script on a Sun SPARC 1+ which has |> version 1.37 of the GNU C compiler/preprocessor |> (cccp and cpp) installed along with the |> vanilla Sun Berkeley C compiler. |> when I ran the Sun C preprocessor (/lib/cpp) I got |> -- |> Chuck Herrick cnh5730@calvin.tamu.edu I posted something simular to this a while back, because I had the same problem (actually I reported it as a bug to NeXT and GNU). The reply I received from RMS was that the behavior of gcc (and hence cc on the NeXT) was the "proper" behavior and that nothing was going to change. The fact that I could produce what *I* thought was the"proper" behavior on an HP (HP-UX 7.0), Sun Sparc (SunOS 4.01), a VAX (BSD 4.3 tahoe), I was a little disappointed in RMS's responce. [Note that none of these C compilers are ANSI compliant, except for HP's with the -Aa switch] So, I posted to comp.lang.c to see what the concensus was there. There answer turned out to be that ANSI compilant preprocessors *ARE REQUIRED* to add the additional space around these tokens. IOW, the Configure script form patch will not run with *ANY* ANSI C. Therefore, the "bug" is with patch, and *NOT* gcc (which is an ANSI compliant compiler). [BTW, when I used the -Aa switch on HP's 7.0 compiler, Configure failed] I sent the whole conversation onto Larry Wall, and he said that in the next distribution of patch, the problem will be fixed. Paul Biron pbiron@ucsd.edu (619) 534-5758 Central University Library, Mail Code C-075-R Social Sciences DataBase Project University of California, San Diego, La Jolla, Ca. 92093