msLoadMap does not properly close map files on parse error

[scott.e@…]

We use C# mapscript on Windows and have been seeing behavior where a map file is
locked by the process that calls msLoadMap.  Thus the file cannot be delete
until the process is finished.  This is only occurs on parse error, and is
because fclose() is not properly called in these circumstances.

Attached is a patch which rectifies this by calling fclose() in all circumstances.

[scott.e@…]

patch

[scott.e@…]

corrected patch

[sdlime]

Will apply at first opportunity...

Steve

[sdlime]

Scott: Applying the patch failed for me for both 4.10 and CVS HEAD. What version
are you using?

Steve

[scott.e@…]

newlines fixed patch

[sdlime]

There's still a problem. Looks like the patch contains some extraneous stuff
that must be custom for you. The patch applies ok but won't compile (against CVS
HEAD). I'm getting:

mapfile.c:2636: error: `MS_SQLSPATIAL' undeclared (first use in this function)
mapfile.c:2636: error: (Each undeclared identifier is reported only once
mapfile.c:2636: error: for each function it appears in.)
mapfile.c: In function `loadMapInternal':
mapfile.c:4571: warning: unused variable `i'
mapfile.c:4571: warning: unused variable `j'
mapfile.c:4571: warning: unused variable `k'
mapfile.c:4572: warning: unused variable `szPath'
mapfile.c:4572: warning: unused variable `szCWDPath'
mapfile.c: In function `ParseMap':
mapfile.c:4604: error: argument `filename' doesn't match prototype
mapfile.c:229: error: prototype declaration
mapfile.c:4604: error: argument `new_mappath' doesn't match prototype
mapfile.c:229: error: prototype declaration
mapfile.c:4629: warning: passing arg 1 of `strdup' makes pointer from integer
without a cast
mapfile.c:4631: warning: passing arg 1 of `getPath' makes pointer from integer
without a cast
mapfile.c: In function `msLoadMap':
mapfile.c:4835: warning: unused variable `starttime'
mapfile.c:4835: warning: unused variable `endtime'

Steve

[scott.e@…]

strip custom stuff

[hobu]

Steve,

What's the status on this one? I know you've been reworking a lot of things for the 5.0 release, and I was wondering if these patches or something like it have made it into trunk yet.

Howard

[hobu]

r6197 and r6200 are a few changesets attempting to reconcile this issue. More changes likely forthcoming for FONTSET and SYMBOLs.

[sdlime]

I've no clue if this is still an issue. I'd love to close if anyone knows definitively...

Steve