[Issue 885] New: Recursion into folders with FindFirstFileA/FindNextFileA behaves incorrectly

[kris]

d-bugmail at puremagic.com wrote:
> http://d.puremagic.com/issues/show_bug.cgi?id=885
> 
>            Summary: Recursion into folders with FindFirstFileA/FindNextFileA
>                     behaves incorrectly
>            Product: D
>            Version: 1.001
>           Platform: PC
>         OS/Version: Windows
>             Status: NEW
>           Severity: major
>           Priority: P2
>          Component: www.digitalmars.com
>         AssignedTo: bugzilla at digitalmars.com
>         ReportedBy: fdp at mcs.be
> 
> 
> The accompanying program was used to find files in a (quite huge) directory
> structure using a regex expression.  When I noticed that some files that were
> there were not found, I added a line to print the directories that get
> 'visited'.  That's where I noticed that some directories were not recursed
> into.  Furthermore this behaviour is not constant; the directories that are
> skipt are not always the same ones.  Thinking of a memory problem, I also
> experimented by adding fullCollect and genCollect calls at various places and
> this had indeed an influence on which directories were skipped.
> 
> Here is the code:
> 
> 
> import std.c.windows.windows;
> import std.path;
> import std.stdio;
> import std.regexp;
> import std.utf;
> 
> void visit (char[] p_strFolder, RegExp p_patt)
> {
>    writefln (">>> ", p_strFolder); // added for debugging purposes
>    WIN32_FIND_DATA fd;
>    char[] strSpec = join (p_strFolder, "*");
>    HANDLE h = FindFirstFileA (cast(char*)strSpec, &fd);
>    bool bOK = h != INVALID_HANDLE_VALUE;
>    while (bOK)
>    {
>       wchar[] wbuf;
> 
>       size_t clength = std.string.strlen(fd.cFileName.ptr);
>          // Convert cFileName[] to unicode
>       size_t wlength =
> MultiByteToWideChar(0,0,fd.cFileName.ptr,clength,null,0);
>       if (wlength > wbuf.length)
>       {
>          wbuf.length = wlength;
>       }
>       size_t n =
> MultiByteToWideChar(0,0,fd.cFileName.ptr,clength,cast(wchar*)wbuf,wlength);
>       assert(n == wlength);
>          // toUTF8() returns a new buffer
>       char[] strName = std.utf.toUTF8(wbuf[0 .. wlength]);
> 
>       if ((fd.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY) != 0)
>       {
>          if (0 != std.string.cmp (strName, ".")  &&
>              0 != std.string.cmp (strName, "..") &&
>              0 != std.string.cmp (strName, ".svn"))
>          {
>             visit (join (p_strFolder, strName), p_patt);
>          }
>       }
>       else
>       {
>          char[] strFullName = join (p_strFolder, strName);
>          if (-1 != p_patt.find(std.string.tolower (strFullName)))
>          {
>            // removed code for debugging purposes
>          }
>       }
>       bOK = cast(bool)FindNextFileA (h, &fd);
>    }
>    FindClose (h);
> }
> 
> void main (char [][] p_args)
> {
>    if (p_args.length == 1)
>    {
>       visit (curdir, RegExp ("."));
>    }
>    else
>    {
>       visit (p_args[1], RegExp (std.string.tolower (p_args[1])));
>    }
> }
> 
> PS: I just tested on 1.002 also and the behaviour is the same.
> 
> 

You're certain FindFirst() et al can be recursed like this? IIRC, they 
maintain some global state?