* mdocml: POSIX allows PATH_MAX to not be defined, meaning "unlimited".
@ 2016-08-02 11:10 schwarze
0 siblings, 0 replies; only message in thread
From: schwarze @ 2016-08-02 11:10 UTC (permalink / raw)
To: source
Log Message:
-----------
POSIX allows PATH_MAX to not be defined, meaning "unlimited".
Found by Aaron M. Ucko <amu at alum dot mit dot edu> on the GNU Hurd,
via Bdale Garbee, https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=829624
Also add EFTYPE at two places where it was forgotten.
Modified Files:
--------------
mdocml:
Makefile
TODO
compat_fts.c
configure
configure.local.example
Added Files:
-----------
mdocml:
test-PATH_MAX.c
Revision Data
-------------
Index: configure
===================================================================
RCS file: /home/cvs/mdocml/mdocml/configure,v
retrieving revision 1.48
retrieving revision 1.49
diff -Lconfigure -Lconfigure -u -p -r1.48 -r1.49
--- configure
+++ configure
@@ -58,6 +58,7 @@ HAVE_ISBLANK=
HAVE_MKDTEMP=
HAVE_NANOSLEEP=
HAVE_OHASH=
+HAVE_PATH_MAX=
HAVE_PLEDGE=
HAVE_PROGNAME=
HAVE_REALLOCARRAY=
@@ -203,6 +204,7 @@ runtest getline GETLINE || true
runtest getsubopt GETSUBOPT || true
runtest isblank ISBLANK || true
runtest mkdtemp MKDTEMP || true
+runtest PATH_MAX PATH_MAX || true
runtest pledge PLEDGE || true
runtest sandbox_init SANDBOX_INIT || true
runtest progname PROGNAME || true
@@ -293,6 +295,7 @@ echo "#define MANPATH_DEFAULT \"${MANPAT
[ -n "${UTF8_LOCALE}" ] && echo "#define UTF8_LOCALE \"${UTF8_LOCALE}\""
[ -n "${HOMEBREWDIR}" ] && echo "#define HOMEBREWDIR \"${HOMEBREWDIR}\""
[ ${HAVE_EFTYPE} -eq 0 ] && echo "#define EFTYPE EINVAL"
+[ ${HAVE_PATH_MAX} -eq 0 ] && echo "#define PATH_MAX 4096"
cat << __HEREDOC__
#define HAVE_DIRENT_NAMLEN ${HAVE_DIRENT_NAMLEN}
--- /dev/null
+++ test-PATH_MAX.c
@@ -0,0 +1,30 @@
+/*
+ * POSIX allows PATH_MAX to not be defined, see
+ * http://pubs.opengroup.org/onlinepubs/9699919799/functions/sysconf.html;
+ * the GNU Hurd is an example of a system not having it.
+ *
+ * Arguably, it would be better to test sysconf(_SC_PATH_MAX),
+ * but since the individual *.c files include "config.h" before
+ * <limits.h>, overriding an excessive value of PATH_MAX from
+ * "config.h" is impossible anyway, so for now, the simplest
+ * fix is to provide a value only on systems not having any.
+ * So far, we encountered no system defining PATH_MAX to an
+ * impractically large value, even though POSIX explicitly
+ * allows that.
+ *
+ * The real fix would be to replace all static buffers of size
+ * PATH_MAX by dynamically allocated buffers. But that is
+ * somewhat intrusive because it touches several files and
+ * because it requires changing struct mlink in mandocdb.c.
+ * So i'm postponing that for now.
+ */
+
+#include <limits.h>
+#include <stdio.h>
+
+int
+main(void)
+{
+ printf("PATH_MAX is defined to be %ld\n", (long)PATH_MAX);
+ return 0;
+}
Index: compat_fts.c
===================================================================
RCS file: /home/cvs/mdocml/mdocml/compat_fts.c,v
retrieving revision 1.9
retrieving revision 1.10
diff -Lcompat_fts.c -Lcompat_fts.c -u -p -r1.9 -r1.10
--- compat_fts.c
+++ compat_fts.c
@@ -68,6 +68,9 @@ static unsigned short fts_stat(FTS *, F
#ifndef O_CLOEXEC
#define O_CLOEXEC 0
#endif
+#ifndef PATH_MAX
+#define PATH_MAX 4096
+#endif
#define CLR(opt) (sp->fts_options &= ~(opt))
#define ISSET(opt) (sp->fts_options & (opt))
Index: TODO
===================================================================
RCS file: /home/cvs/mdocml/mdocml/TODO,v
retrieving revision 1.220
retrieving revision 1.221
diff -LTODO -LTODO -u -p -r1.220 -r1.221
--- TODO
+++ TODO
@@ -620,6 +620,13 @@ Several areas can be cleaned up to make
* structural issues
************************************************************************
+- POSIX says in the documentation of sysconf(3) that PATH_MAX
+ is allowed to be so large that it is a bad idea to use it
+ for sizing static buffers. So use dynamic buffers throughout.
+ See the file test-PATH_MAX.c for details.
+ Found by Aaron M. Ucko in the GNU Hurd via Bdale Garbee,
+ https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=829624
+
- We use the input line number at several places to distinguish
same-line from different-line input. That plainly doesn't work
with user-defined macros, leading to random breakage.
Index: Makefile
===================================================================
RCS file: /home/cvs/mdocml/mdocml/Makefile,v
retrieving revision 1.490
retrieving revision 1.491
diff -LMakefile -LMakefile -u -p -r1.490 -r1.491
--- Makefile
+++ Makefile
@@ -20,6 +20,7 @@ VERSION = 1.14.0
# === LIST OF FILES ====================================================
TESTSRCS = test-dirent-namlen.c \
+ test-EFTYPE.c \
test-err.c \
test-fts.c \
test-getline.c \
@@ -28,6 +29,7 @@ TESTSRCS = test-dirent-namlen.c \
test-mkdtemp.c \
test-nanosleep.c \
test-ohash.c \
+ test-PATH_MAX.c \
test-pledge.c \
test-progname.c \
test-reallocarray.c \
Index: configure.local.example
===================================================================
RCS file: /home/cvs/mdocml/mdocml/configure.local.example,v
retrieving revision 1.18
retrieving revision 1.19
diff -Lconfigure.local.example -Lconfigure.local.example -u -p -r1.18 -r1.19
--- configure.local.example
+++ configure.local.example
@@ -227,6 +227,7 @@ CFLAGS="-g"
# be regarded as successful).
HAVE_DIRENT_NAMLEN=0
+HAVE_EFTYPE=0
HAVE_ERR=0
HAVE_FTS=0
HAVE_GETLINE=0
@@ -234,6 +235,7 @@ HAVE_GETSUBOPT=0
HAVE_ISBLANK=0
HAVE_MKDTEMP=0
HAVE_OHASH=0
+HAVE_PATH_MAX=0
HAVE_PLEDGE=0
HAVE_PROGNAME=0
HAVE_REALLOCARRAY=0
--
To unsubscribe send an email to source+unsubscribe@mdocml.bsd.lv
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2016-08-02 11:10 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-08-02 11:10 mdocml: POSIX allows PATH_MAX to not be defined, meaning "unlimited" schwarze
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).