Fix shell handling of escapes

re: https://github.com/Unidata/netcdf-c/issues/1988

There was an issue with certain shell programs (bash notably).
For certain platforms and when given a url that had an escaped
'#' character (e.g. \\#) bash would not remove the backslash. So I
had to add a hack for this. Unfortunately I overdid it and it
removed all '' characters. This is ok for non-windows platforms,
but obviously fails for windows.

The fix is this.

1. In a utility program (ncgen, ncdump, nccopy, etc) there is probably a call (or calls) to NC_backslashUnescape(xxx) where xxx is a path argument from the command line.
2. Replace each such call with NC_shellUnescape(xxx).

The NC_shellUnescape function was added and searched only for occurrences of "\#" and replaces them with "#".

include/ncpathmgr.h

@@ -168,5 +168,6 @@ EXTERNL void printutf8hex(const char* s, char* sx);
 /* From dutil.c */
 EXTERNL char* NC_backslashEscape(const char* s);
 EXTERNL char* NC_backslashUnescape(const char* esc);
+EXTERNL char* NC_shellUnescape(const char* esc);
 
 #endif /* _NCPATHMGR_H_ */

libdispatch/dutil.c

@@ -164,6 +164,37 @@ NC_entityescape(const char* s)
     return escaped;
 }
 
+char*
+/*
+Depending on the platform, the shell will sometimes
+pass an escaped octotherpe character without removing
+the backslash. So this function is appropriate to be called
+on possible url paths to unescape such cases. See e.g. ncgen.
+*/
+NC_shellUnescape(const char* esc)
+{
+    size_t len;
+    char* s;
+    const char* p;
+    char* q;
+
+    if(esc == NULL) return NULL;
+    len = strlen(esc);
+    s = (char*)malloc(len+1);
+    if(s == NULL) return NULL;
+    for(p=esc,q=s;*p;) {
+	switch (*p) {
+	case '\\':
+	     if(p[1] == '#')
+	         p++;
+	     /* fall thru */
+	default: *q++ = *p++; break;
+	}
+    }
+    *q = '\0';
+    return s;
+}
+
 /**
 Wrap mktmp and return the generated path,
 or null if failed.
@@ -371,3 +402,4 @@ int isnan(double x)
 }
 
 #endif /*APPLE*/
+

ncdump/nccopy.c

@@ -2412,8 +2412,8 @@ main(int argc, char**argv)
 	error("one input file and one output file required");
     }
     /* Canonicalize the input and output files names */
-    inputfile = NC_backslashUnescape(argv[0]); /* Remove shell added escapes */
-    outputfile = NC_backslashUnescape(argv[1]);
+    inputfile = NC_shellUnescape(argv[0]); /* Remove shell added escapes */
+    outputfile = NC_shellUnescape(argv[1]);
     if(strcmp(inputfile, outputfile) == 0) {
 	error("output would overwrite input");
     }

ncdump/ncdump.c

@@ -2335,7 +2335,7 @@ main(int argc, char *argv[])
 
     /* We need to look for escape characters because the argument
        may have come in via a shell script */
-    path = NC_backslashUnescape(argv[i]);
+    path = NC_shellUnescape(argv[i]);
     if(path == NULL) {
 	snprintf(errmsg,sizeof(errmsg),"out of memory un-escaping argument %s", argv[i]);
 	goto fail;

ncgen/main.c

@@ -351,7 +351,7 @@ main(
 	  break;
 	case 'o':		/* to explicitly specify output name */
 	  if(netcdf_name) efree(netcdf_name);
-	  netcdf_name = NC_backslashUnescape(optarg);
+	  netcdf_name = NC_shellUnescape(optarg);
 	  break;
 	case 'P': /* diskless with persistence */
 	  diskless = 1;
@@ -460,7 +460,7 @@ main(
 	}
     }
 
-    cdlname = NC_backslashUnescape(argv[0]);
+    cdlname = NC_shellUnescape(argv[0]);
     if(cdlname != NULL) {
 	if(strlen(cdlname) > NC_MAX_NAME)
 	  cdlname[NC_MAX_NAME] = '\0';

nczarr_test/ncdumpchunks.c

@@ -485,7 +485,7 @@ main(int argc, char** argv)
     }
 
     {
-        char* s = NC_backslashUnescape(argv[0]);
+        char* s = NC_shellUnescape(argv[0]);
 	strcpy(format.file_name,filenamefor(s));
 	nullfree(s);
     }

nczarr_test/s3util.c

@@ -163,7 +163,7 @@ main(int argc, char** argv)
 	    dumpoptions.nctype = typefor(optarg);
 	    break;
 	case 'u': {
- 	        char* p = NC_backslashUnescape(optarg);
+ 	        char* p = NC_shellUnescape(optarg);
 		ncuriparse(p,&dumpoptions.url);
 		nullfree(p);
 		if(dumpoptions.url == NULL) {

nczarr_test/tst_utils.c

@@ -129,7 +129,7 @@ getoptions(int* argcp, char*** argvp)
     *argvp += optind;
 
     if(*argcp > 0) {
-	char* p = NC_backslashUnescape((*argvp)[0]);
+	char* p = NC_shellUnescape((*argvp)[0]);
         strcpy(options->file,filenamefor(p));
 	nullfree(p);
     }

nczarr_test/zmapio.c

@@ -177,7 +177,7 @@ main(int argc, char** argv)
     }
 
     {
-        char* p = NC_backslashUnescape(argv[0]);
+        char* p = NC_shellUnescape(argv[0]);
         strcpy(dumpoptions.infile,filenamefor(p));
 	if(p) free(p);
     }

unit_test/test_pathcvt.c

@@ -27,6 +27,7 @@ typedef struct Test {
 static Test PATHTESTS[] = {
 {"/xxx/a/b",{"/xxx/a/b", "/c/xxx/a/b", "/cygdrive/c/xxx/a/b", "c:\\xxx\\a\\b"}},
 {"d:/x/y",{ "/d/x/y", "/d/x/y",  "/cygdrive/d/x/y",  "d:\\x\\y"}},
+{"d:\\x\\y",{ "/d/x/y", "/d/x/y",  "/cygdrive/d/x/y",  "d:\\x\\y"}},
 {"/cygdrive/d/x/y",{ "/d/x/y", "/d/x/y", "/cygdrive/d/x/y",  "d:\\x\\y"}},
 {"/d/x/y",{ "/d/x/y", "/d/x/y",  "/cygdrive/d/x/y",  "d:\\x\\y"}},
 {"/cygdrive/d",{ "/d", "/d",  "/cygdrive/d",  "d:"}},
@@ -38,6 +39,7 @@ static Test PATHTESTS[] = {
     "d:\\git\\netcdf-c\\dap4_test\\test_anon_dim.2.syn"}},
 /* Test relative path */
 {"x/y",{ "x/y", "x/y", "x/y",  "x\\y"}},
+{"x\\y",{ "x/y", "x/y", "x/y",  "x\\y"}},
 #ifndef _WIN32
 /* Test utf8 path */
 {"/海/海",{ "/海/海", "/c/海/海", "/cygdrive/c/海/海",  "c:\\海\\海"}},
@@ -54,6 +56,7 @@ main(int argc, char** argv)
     Test* test;
     int failcount = 0;
     char* cvt = NULL;
+    char* unescaped = NULL;
     int k;
     int drive = 'c';
 
@@ -71,7 +74,9 @@ main(int argc, char** argv)
 #endif
 	        continue;
 	    }
-   	    cvt = NCpathcvt_test(test->test,kind,drive);
+	    /* ensure that NC_shellUnescape does not affect result */
+	    unescaped = NC_shellUnescape(test->test);	
+   	    cvt = NCpathcvt_test(unescaped,kind,drive);
 #ifdef DEBUG
 	    fprintf(stderr,"TEST local=%s: input: |%s| expected=|%s| actual=|%s|: ",
 			kind2string(kind),test->test,test->expected[k],cvt);
@@ -95,10 +100,11 @@ main(int argc, char** argv)
 #ifdef DEBUG
 	    fprintf(stderr,"\n");
 #endif	    
+	    nullfree(unescaped); unescaped = NULL;
 	    nullfree( cvt); cvt = NULL;
 	}
     }
-    nullfree(cvt);
+    nullfree(cvt); nullfree(unescaped);
     fprintf(stderr,"%s test_pathcvt\n",failcount > 0 ? "***FAIL":"***PASS");
     nc_finalize();
     return (failcount > 0 ? 1 : 0);