add man page for scan_base64, handle partial blocks better

textcode/scan_base64.3

@@ -0,0 +1,34 @@
+.TH scan_base64 3
+.SH NAME
+scan_base64 \- decode base64 encoded data
+.SH SYNTAX
+.B #include <textcode.h>
+
+size_t \fBscan_base64\fP(const char *\fIsrc\fR,char *\fIdest\fR,size_t* \fIdestlen\fR);
+
+.SH DESCRIPTION
+scan_base64 decodes base64 encoded data from src into dest.
+It will stop when it encountes any non-valid input characters.
+It will then write the number of decoded bytes in dest into *destlen,
+and return the number of bytes decoded from src.
+
+Note that real world base64 encoded data is sometimes permitted to
+contain whitespace characters or new lines. This function will not allow
+those and return the decoded data until then.
+
+base64 works by taking 3 bytes of binary input and converting them into
+4 bytes of printable ASCII. If the input ends in the middle of a base64
+4-byte-tuple, scan_base64 will disregard it.
+
+Many base64 variants demand padding in the last block. Some don't. This
+implementation will consume padding if it is there, but will not
+complain if it is not.
+
+.SH "RETURN VALUE"
+scan_base64 returns the number of bytes successfully scanned and
+processed from src.
+.SH EXAMPLES
+scan_base64("Zm5vcmQ=",buf,&i) -> return 8, i=5, buf="fnord"
+
+.SH "SEE ALSO"
+scan_xlong(3), scan_8long(3), fmt_ulong(3)

textcode/scan_base64.c

@@ -15,20 +15,39 @@ static inline int dec(unsigned char x) {
 size_t scan_base64(const char *src,char *dest,size_t *destlen) {
   unsigned short tmp=0,bits=0;
   register const unsigned char* s=(const unsigned char*) src;
-  const char* orig=dest;
-  for (;;) {
+  const char* orig=src;
+  size_t i,j=0;
+  for (i=0;;) {
     int a=dec(*s);
     if (a<0) {
+      int equal=(*s=='=');
       while (*s=='=' && ((s-(const unsigned char*)src)&3)) ++s;
+      tmp &= ((1<<bits)-1);
+      if (!tmp || equal) { j=i; orig=s; }
       break;
     }
     tmp=(tmp<<6)|a; bits+=6;
     ++s;
     if (bits>=8) {
-      *dest=(tmp>>(bits-=8));
-      ++dest;
+      bits-=8;
+      if (dest) dest[i]=(tmp>>bits);
+      ++i;
+      if (!bits) { j=i; orig=s; }
     }
   }
-  *destlen=dest-orig;
-  return (const char*)s-src;
+  if (destlen) *destlen=j;
+  return orig-src;
 }
+
+#ifdef UNITTEST
+#include <assert.h>
+#include <string.h>
+#include <stdio.h>
+
+int main() {
+  char buf[100];
+  size_t i,l;
+  memset(buf,0,10); assert(scan_base64("Zm5vcmQ=",buf,&l)==8 && l==5 && !memcmp(buf,"fnord",6));
+  memset(buf,0,10); assert(scan_base64("Zm5vcmQ",buf,&l)==7 && l==5 && !memcmp(buf,"fnord",6));
+}
+#endif