MinGW compilation for 0.04

[roytam1]

this patch fix 2 issues:

• the use of vasprintf()
• ignoring(treat as successful) if mkdir sets errno EEXIST which means directory exists

index c01b3a6..b0e09ae 100644
--- a/lib/lha_arch_win32.c
+++ b/lib/lha_arch_win32.c
@@ -32,10 +32,16 @@ CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 #include <utime.h>
 #include <sys/stat.h>
 #include <sys/types.h>
+#include <errno.h>

 int lha_arch_mkdir(char *path, unsigned int unix_mode)
 {
-       return mkdir(path) == 0;
+       int i;
+
+       i = mkdir(path);
+       if(i < 0 && errno == EEXIST) /* ignore if dir exists */
+               i = 0;
+       return i == 0;
 }

 int lha_arch_chown(char *filename, int unix_uid, int unix_gid)
diff --git a/src/safe.c b/src/safe.c
index 9e3034c..373033f 100644
--- a/src/safe.c
+++ b/src/safe.c
@@ -23,6 +23,18 @@ CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 #include <stdarg.h>
 #include <stdlib.h>

+#if LHA_ARCH == LHA_ARCH_WINDOWS
+#include <stdarg.h>
+int vasprintf( char **sptr, char *fmt, va_list argv )
+{
+    int wanted = vsnprintf( *sptr = NULL, 0, fmt, argv );
+    if( (wanted > 0) && ((*sptr = malloc( 1 + wanted )) != NULL) )
+        return vsprintf( *sptr, fmt, argv );
+
+    return wanted;
+}
+#endif
+
 // Routines for safe terminal output.
 //
 // Data in LHA files (eg. filenames) may contain malicious string

[fragglet]

Thanks for the patch. Would you be willing to rework it slightly?

I've set up the lha_arch layer to separate the architecture-specific code from the non-specific code, to specifically avoid putting "#if LHA_ARCH ==" throughout the program. So if vasprintf is a problem on Windows, it should be moved into the layer.

Also, I'm a bit confused as to why mkdir() needs to ignore existing directories. As far as I know, the extraction code should only be creating directories if they don't already exist. Is this not the case?

Thanks!

[roytam1]

Also, I'm a bit confused as to why mkdir() needs to ignore existing directories. As far as I know, the extraction code should only be creating directories if they don't already exist. Is this not the case?

I didn't dig though the sources, but it creates the directory and it checks the newly created directory every file that inside that directory when extracting.

EDIT:

LHAFileType lha_arch_exists(char *filename)
{
    // TODO
    return LHA_FILE_NONE;
}

so that's the problem.

file: http://homepage3.nifty.com/csdinc/archiver/lib/bga32037.lzh

without the patch:

16:42 D:\>lha bga32037.lzh
 PERMSSN    UID  GID      SIZE  RATIO     STAMP           NAME
---------- ----------- ------- ------ ------------ --------------------
[MS-DOS]                372736  57.1% Jun  4  2002 Bga32.dll
[MS-DOS]                 19586  37.7% Jun  4  2002 Bga32.txt
[MS-DOS]                     0 ****** Jun  4  2002 sdk/
[MS-DOS]                 25338  24.2% Jun  4  2002 SDK/Bga32Api.txt
[MS-DOS]                 14911  33.7% Jun  4  2002 SDK/Bga32Cmd.txt
[MS-DOS]                 17213  29.3% Jun  4  2002 SDK/Bga32Api.h
[MS-DOS]                  6144  20.5% Jun  4  2002 SDK/Bga32.lib
---------- ----------- ------- ------ ------------ --------------------
 Total         7 files  455928  52.2% Oct 18  2010

16:42 D:\>lha x bga32037.lzh
Bga32.dll       - Melted   :  oooooooooooooooooooooooooooooooooooooooooooooo
Bga32.txt       - Melted   :  ooo
Failed to create parent directory SDK/
Failed to create parent directory SDK/
Failed to create parent directory SDK/
Failed to create parent directory SDK/

16:43 D:\>dir/b sdk

with the patch:

16:43 D:\>lha x bga32037.lzh
Bga32.dll       - Melted   :  oooooooooooooooooooooooooooooooooooooooooooooo
Bga32.txt       - Melted   :  ooo
SDK/Bga32Api.txt        - Melted   :  oooo
SDK/Bga32Cmd.txt        - Melted   :  oo
SDK/Bga32Api.h  - Melted   :  ooo
SDK/Bga32.lib   - Melted   :  o

16:44 D:\>dir/b sdk
Bga32.lib
Bga32Api.h
Bga32Api.txt
Bga32Cmd.txt

[roytam1]

Thanks for the patch. Would you be willing to rework it slightly?

Sure.

diff --git a/lib/lha_arch_win32.c b/lib/lha_arch_win32.c
index c01b3a6..15d86e9 100644
--- a/lib/lha_arch_win32.c
+++ b/lib/lha_arch_win32.c
@@ -32,10 +32,26 @@ CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 #include <utime.h>
 #include <sys/stat.h>
 #include <sys/types.h>
+#include <errno.h>
+#include <stdarg.h>
+
+int vasprintf( char **sptr, char *fmt, va_list argv )
+{
+    int wanted = vsnprintf( *sptr = NULL, 0, fmt, argv );
+    if( (wanted > 0) && ((*sptr = malloc( 1 + wanted )) != NULL) )
+        return vsprintf( *sptr, fmt, argv );
+
+    return wanted;
+}

 int lha_arch_mkdir(char *path, unsigned int unix_mode)
 {
-       return mkdir(path) == 0;
+       int i;
+
+       i = mkdir(path);
+       if(i < 0 && errno == EEXIST) /* ignore if dir exists */
+               i = 0;
+       return i == 0;
 }

 int lha_arch_chown(char *filename, int unix_uid, int unix_gid)

[fragglet]

Ah, I see. The problem is with lha_arch_exists, not lha_arch_mkdir.

Several functions in lha_arch_win32.c aren't implemented yet. I'm hoping to put together a fully working and tested MingW build soon - as you can see it's rather sketchy at the moment.

[fragglet]

Hi,

I've implemented the incomplete functions in lha_arch_win32.c and got the test suite running on Windows. The current version on head passes all tests, so hopefully it should work a bit better for you.

Thanks!