From 1fb94211638fdf7bf8187b536a9c638600ddc3fe Mon Sep 17 00:00:00 2001
From: Jean-Francois Dockes <jf@dockes.org>
Date: Fri, 28 Feb 2020 09:22:03 +0100
Subject: [PATCH] OCR: small adjustments for Windows

---
 src/filters/rclocr.py                | 27 +++++++++++++++++++++++++--
 src/filters/rclpdf.py                |  3 ++-
 src/python/recoll/recoll/conftree.py | 20 ++++++++++----------
 3 files changed, 37 insertions(+), 13 deletions(-)

diff --git a/src/filters/rclocr.py b/src/filters/rclocr.py
index c6b17165..9bcb1635 100755
--- a/src/filters/rclocr.py
+++ b/src/filters/rclocr.py
@@ -31,13 +31,32 @@ import importlib.util
 import rclconfig
 import rclocrcache
 
+_mswindows = (sys.platform == "win32")
 def _deb(s):
-    print("rclocr: %s" % s, file=sys.stderr)
+    if not _mswindows:
+        print("rclocr: %s" % s, file=sys.stderr)
     
 def Usage():
     _deb("Usage: rclocr.py <imagefilename>")
     sys.exit(1)
 
+def breakwrite(f, data):
+    # On Windows, writing big chunks can fail with a "not enough space"
+    # error. Seems a combined windows/python bug, depending on versions.
+    # See https://bugs.python.org/issue11395
+    # In any case, just break it up
+    total = len(data)
+    bs = 4*1024
+    offset = 0
+    while total > 0:
+        if total < bs:
+            tow = total
+        else:
+            tow = bs
+        f.write(data[offset:offset+tow])
+        offset += tow
+        total -= tow
+
 if len(sys.argv) != 2:
     Usage()
 
@@ -50,7 +69,11 @@ cache = rclocrcache.OCRCache(config)
 
 incache, data = cache.get(path)
 if incache:
-    sys.stdout.buffer.write(data)
+    try:
+        breakwrite(sys.stdout.buffer, data)
+    except Exception as e:
+        _deb("RCLOCR error writing: %s" % e)
+        sys.exit(1)
     sys.exit(0)
     
 #### Data not in cache
diff --git a/src/filters/rclpdf.py b/src/filters/rclpdf.py
index 27fc339a..830b8e6c 100755
--- a/src/filters/rclpdf.py
+++ b/src/filters/rclpdf.py
@@ -410,7 +410,8 @@ class PDFExtractor:
                            self.filename]
                     data = subprocess.check_output(cmd)
                     html = _htmlprefix + self.em.htmlescape(data) + _htmlsuffix
-                except:
+                except Exception as e:
+                    self.em.rclog("%s failed: %s" % (cmd, e))
                     pass
 
         if self.extrameta:
diff --git a/src/python/recoll/recoll/conftree.py b/src/python/recoll/recoll/conftree.py
index 99e6eceb..53d445c7 100644
--- a/src/python/recoll/recoll/conftree.py
+++ b/src/python/recoll/recoll/conftree.py
@@ -195,22 +195,22 @@ class ConfTree(ConfSimple):
             raise TypeError("getbin: parameters must be binary not unicode")
         #_debug("ConfTree::getbin: nm [%s] sk [%s]" % (nm, sk))
         
-        if sk == b'' or sk[0] != b'/'[0]:
+        # Note the test for root. There does not seem to be a direct
+        # way to do this in os.path
+        if not sk:
             return ConfSimple.getbin(self, nm, sk)
 
-        if sk[len(sk)-1] == b'/'[0]:
-             sk = sk[:len(sk)-1]
-
         # Try all sk ancestors as submaps (/a/b/c-> /a/b/c, /a/b, /a, b'')
-        while sk:
+        while True:
             if sk in self.submaps:
                 return ConfSimple.getbin(self, nm, sk)
             if sk + b'/' in self.submaps:
-                return ConfSimple.getbin(self, nm, sk+b'/')
-            i = sk.rfind(b'/')
-            if i == -1:
-                break
-            sk = sk[:i]
+                return ConfSimple.getbin(self, nm, sk + b'/')
+            nsk = os.path.dirname(sk)
+            if nsk == sk:
+                # sk was already root, we're done. 
+                break;
+            sk = nsk
 
         return ConfSimple.getbin(self, nm)