Thread

1. isolationtester and 'specs' subdirectory

   Josh Kupershmidt <schmiddy@gmail.com> — 2013-06-20T23:10:30Z

   Hi all,
   I have a Debian machine with gcc 4.7.2-5 where make check-world fails
   in the isolation check, like so:
   
   ...
   make[2]: Leaving directory `/home/josh/src/postgresql/src/test/regress'
   make -C isolation check
   
   [snip]
   
   gcc -O2 -Wall -Wmissing-prototypes -Wpointer-arith
   -Wdeclaration-after-statement -Wendif-labels
   -Wmissing-format-attribute -Wformat-security -fno-strict-aliasing
   -fwrapv -fexcess-precision=standard -g -I.
   -I../../../src/interfaces/libpq -I./../regress -I../../../src/include
   -D_GNU_SOURCE   -c -o isolation_main.o isolation_main.c
   gcc: error: ./specs: Is a directory
   make[2]: *** [isolation_main.o] Error 1
   ...
   
   I eventually tracked down the cause of this failure to a trailing ':'
   in my $LIBRARY_PATH, which causes gcc to look inside the current
   directory for a 'specs' file [1] among other things. Although I
   probably don't need that trailing ':', it seems like we should avoid
   naming this directory 'specs' nonetheless to avoid confusion with gcc.
   
   Renaming the 'specs' directory to something like 'isolation_specs' and
   adjusting isolation_main.c accordingly lets me pass `make
   check-world`. Proposed patch attached.
   
   Josh
   
   [1] http://gcc.gnu.org/ml/gcc-help/2010-05/msg00292.html
   

2. Re: isolationtester and 'specs' subdirectory

   Robert Haas <robertmhaas@gmail.com> — 2013-06-25T16:15:01Z

   On Thu, Jun 20, 2013 at 7:10 PM, Josh Kupershmidt <schmiddy@gmail.com> wrote:
   > I eventually tracked down the cause of this failure to a trailing ':'
   > in my $LIBRARY_PATH, which causes gcc to look inside the current
   > directory for a 'specs' file [1] among other things. Although I
   > probably don't need that trailing ':', it seems like we should avoid
   > naming this directory 'specs' nonetheless to avoid confusion with gcc.
   >
   > Renaming the 'specs' directory to something like 'isolation_specs' and
   > adjusting isolation_main.c accordingly lets me pass `make
   > check-world`. Proposed patch attached.
   
   This seems like pretty stupid behavior on the part of gcc.  And, we're
   generally reluctant to rename things too blithely because it
   complicates back-patching.  But on the flip side, back-patching
   changes to the isolation specs is probably a rare event.
   
   -- 
   Robert Haas
   EnterpriseDB: http://www.enterprisedb.com
   The Enterprise PostgreSQL Company