hook_init called during drush, disables module and throws error, preventing drush further execution

Thank you for the analysis and patch!

Something is wrong with the patch -- it wants to replace the entire file, not just the if. Maybe it's because you changed the file mode?

diff --git a/mailmime.module b/mailmime.module
old mode 100644
new mode 100755

Let's see what the testbot says. Patches need to go against HEAD, i.e. the -dev version.

While you're at it, please make your comment conform to the coding guidelines (space, initial cap, ending dot).

Please set NR to trigger the testbot.

Now it passes, but this is one nasty patch. I don't want to lose the entire history of the file.

There is a super quick and easy fix though, preventing the module disabling code from running during drush

What exactly is happening that disables the module? Can't we be more precise?

(BTW, Drupal coding conventions require ending comments with a dot.)

Ah, I only saw the change in mailmime.inc...

But now that I'm seeing the part in mailmime.module, I'm wondering why the former is needed at all.

+++ b/mailmime.module
@@ -12,52 +12,56 @@
+  /* Prevent module from disabling itself during drush execution,
+  module has all needed requirements but was still disabling itself. */
+  if (!function_exists('drush_main')) {
+    // This only needs to succeed once.

Why not just something more reasonable like

+
+  /* Prevent module from disabling itself during drush execution,
+     module has all needed requirements but was still disabling itself. */
+  if (function_exists('drush_main')) {
+    return;
+  }
+
   // This only needs to succeed once.

?

(Note the white space.)

Why are you still nuking mailmime.inc?

Please be more specific.

The problem with this patch is that it's 58K because of indenting changes, and
a) it's impossible to find the actual changes and
b) Git will mark most everything as changed when it really isn't.

I'm just trying to help out here because the author/maintainer has vanished, and I don't want to commit a patch that destroys the history of the module.

Here's a more reasonable version of #9 that does the same thing except for changing indentation.

You also tried to fix some ugly indentation and line wrapping issues, but we must never fix functionality and hide (!) unrelated cosmetics in the same patch. Except maybe adding an empty line or two...

I got the error, which does not appear when I remove this patch from code.

Fatal error: Call to undefined function url_to_realpath() in /var/www/avtoms/sites/all/modules/mailmime/mailmime.inc on line 393

It should be

if (function_exists('drush_main')) {
  // Prevent module execution during drush execution (#2427405).
  return;
}

instead of

if (!function_exists('drush_main')) {
  // Prevent module execution during drush execution (#2427405).
  return;
}

Otherwhise it will..

• not fix the Drush issue
• break htmlmail module which expects the class MailMIME to exist

So, is this what we need?

After reading again the whole issue and thinking about it, I'd even say that patch from #22 is the correct one.

In #17 you committed a patch which checks for a drush instance in hook_init() as well as in mimemail.inc.

Since the original issue was just about the issue, that drush automatically disables mailmime due to hook_init() execution, I don't see a need to cancel execution in mimemail.inc. It can't hurt drush to have the MailMIME class available.

Indeed, I needed to apply patch #22 yesterday to make cron via drush working again. It's 100% save because the Drush condition in mailmime.inc makes absolutely no sense.

#22 does not apply to the HEAD of 7.x-2.x because the condition is reversed.

So, could this be what we need?

Yes, that's it! Thanks.

Ok, I take this as RTBC...

Thanks!