xarcade2jstick opens the wrong FD

[univerio]

On my system the board shows up as two distinct devices under /dev/input. Due to precarious circumstances, they are /dev/input/event9 and /dev/input/event10, but because xarcade2jstick simply opens the first one it finds and glob(3) returns the entries in lexicographical order, xarcade2jstick opens /dev/input/event10 instead of the correct one, /dev/input/event9. This (event9 and event10) is the only case where it happens because the numbers straddle the single- and double-digit boundary. (I suppose it also happens at event99 and event100.)

Operating under the assumption that the lower number is better, I've patched the driver to sort the entries based on the number. This may not be a correct assumption, which is why I've created this as an issue instead of a PR. The patch follows:

diff --git a/src/input_xarcade.c b/src/input_xarcade.c
index 795edc5..19cf6b0 100755
--- a/src/input_xarcade.c
+++ b/src/input_xarcade.c
@@ -16,6 +16,7 @@
 /*                 Copyright (c) 2014, Florian Mueller                      */
 /* ======================================================================== */

+#include <stdlib.h>
 #include <glob.h>
 #include <errno.h>
 #include "input_xarcade.h"
@@ -57,6 +58,24 @@ int16_t input_xarcade_close(INP_XARC_DEV* const xdev) {

 // supplementary functions -------------------

+static int compareGlobEntry(const void *e1, const void * e2) {
+       const char *name1 = *(const char **)e1;
+       const char *name2 = *(const char **)e2;
+       char *end1, *end2;
+       long i1, i2;
+
+       name1 += 16; // skip over /dev/input/event
+       name2 += 16;
+       end1 = end2 = NULL;
+       i1 = strtol(name1, &end1, 10);
+       i2 = strtol(name2, &end2, 10);
+
+       if (i1 == i2) {
+               return strcmp(end1, end2);
+       }
+       return i1 < i2 ? -1 : 1;
+}
+
 int findXarcadeDevice(void) {
        char name[256];
        char *filename;
@@ -65,12 +84,15 @@ int findXarcadeDevice(void) {
        int rc;
        glob_t pglob;

-       rc = glob("/dev/input/event*", 0, NULL, &pglob);
+       rc = glob("/dev/input/event*", GLOB_NOSORT, NULL, &pglob);
        if (rc) {
                printf("Failed to open event devices\n");
                return -1;
        }

+       // sort matched filenames
+       qsort(pglob.gl_pathv, pglob.gl_pathc, sizeof(char *), compareGlobEntry);
+
        for (ctr = 0; ctr < pglob.gl_pathc; ++ctr) {
                filename = pglob.gl_pathv[ctr];
                fevdev = open(filename, O_RDONLY);

[petrockblog]

Would you say that this patch should be integrated into the master branch?

[univerio]

I think so. It's slightly more brittle due to having to parse the number out of the filename, but it's worth fixing this bug. Although, I don't see anybody else with the same issue, so maybe it's not so important after all.

…

On Wed, Mar 13, 2019, 11:32 PM Florian ***@***.***> wrote: Would you say that this patch should be integrated into the master branch? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#39 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA_TdJx2zPcP6USaxN35pUEf63GUvp5qks5vWezrgaJpZM4J9e9z> .

[xnaron]

Edit: I applied the patch above and now it selects the "correct for me" input.

I have the same issue. Was this patched?

(event9 is correct) but picks 12

[petrockblog]

It would be great if someone could create a PR for the patch above.