13001[BUGFIX] Remove debug statement for not existing classes
Commit message
Review
-
Code review CR +2
- Christian Kuhn +2
- Sebastian Michaelsen +1
- Oliver Klee +1
- Helmut Hummel +1
-
Verified V +2
- Christian Kuhn +2
- Sebastian Michaelsen +1
- Helmut Hummel +1
Files 1
Review log 30
Show 5 earlier messages
-
Markus Klein commented on patch set 1 07/24/12, 17:53
I would prefer that you didn't submit thisclass.t3lib_div.php:4310Adapt these two lines and add: Not used anymore since 6.0 -
Sebastian Michaelsen commented on patch set 1 07/24/12, 17:53
Looks good to me, but someone else must approve Ok, that's a pragmatic solution. Do we have a standard procedure to deprecate parameters? You could write to deprecation log, when the $silent-Parameter is used. -
Mario Rimann commented on patch set 2 V 0 07/24/12, 17:53
I've just tested it -> the error messages (debug output) are gone, nice! -
Markus Klein commented on patch set 2 07/24/12, 17:53
Looks good to me, but someone else must approve -
Oliver Klee commented on patch set 2 07/24/12, 17:53
Looks good to me, but someone else must approve
-
Christian Kuhn commented on patch set 2 07/24/12, 17:53
seems to only fix the consequence, maybe we need to check all class_exist calls. Will give a real review tomorrow. -
Helmut Hummel commented on patch set 3 07/24/12, 17:53
I would definitely go through all the places that call getUserObj with a non existing classname and fix these places, too. Does not make sense to call that function with sth. like "admin" or "required" -
Helmut Hummel commented on patch set 3 07/24/12, 17:53
Hm, although I do not like the approach of silent fails, I fear that the exact purpose of getUserObject is to have exactly that. How should we go on with it? Options: 1. Merge this patch and leave the core like this 2. Wrap all getUserObject calls in the core with class_exists() and throw a deprecation message when the class is not found 3. Wrap all getUserObject calls in the core with class_exists() and throw an exception if the class is not found I think 3. would be too obtrusive, so let's pick 1. or 2. I'm undecided... -
Helmut Hummel commented on patch set 3 V 0 07/24/12, 17:53
by team member; Looks good to me, approved In the long run, we should deprecate: 1. The pseudo singleton 2. The usage of a filename to require it (we have an autoloader) 3. Then we can deprecate the whole method. All these things are not scope of this patch so go ahead and merge it. We cannot do something else than that without deprecating first. -
Helmut Hummel commented on patch set 3 07/24/12, 17:53
@Sebastian: In general I totally agree. Unfortunately just a class_exists is not enough because the class might be required by getUserObj. So there's no way around living with the silent failure for now and deprecate this method altogether in the long run -
Helmut Hummel uploaded patch set 4 07/24/12, 17:53
-
Helmut Hummel commented on patch set 4 07/24/12, 17:53
Verified; Looks good to me, but someone else must approve @Sebastian: Right. Is also the scope of a different patch. There are a lot of array key usages in the core without a check. -
Oliver Klee commented on patch set 5 07/24/12, 17:53
Looks good to me, but someone else must approve -
Helmut Hummel uploaded patch set 5 07/24/12, 17:53
-
Helmut Hummel commented on patch set 5 07/24/12, 17:53
Verified; Looks good to me, but someone else must approve -
Christian Kuhn commented on patch set 5 V 0 07/24/12, 17:53
by team member; Looks good to me, approved Went through the patch and the one that introduced this regression. I must say that I didn't expect that getUserObj will be called with such a broken config (like '&required'), we should at least sanitize the calling methods in this regard, but we must also keep compatibility with the old behavior. So, a patch for getUserObj is needed anyway, and just removing the debug output seems to be the most clever strategy for now. We should have the method on our radar and maybe improve the handling in stabilization phase again. For now, I'm +1 for the patch and will merge it now. -
Georg Ringer commented on patch set 3 07/24/12, 17:53
Verified; Looks good to me, but someone else must approve just fixed a typo in the commit msg (therefor > therefore), so all votes are still valid -
Georg Ringer uploaded patch set 3 07/24/12, 17:53
-
Sebastian Michaelsen commented on patch set 2 07/24/12, 17:53
Verified; Looks good to me, but someone else must approve -
Sebastian Michaelsen commented on patch set 3 07/24/12, 17:53
I'd also be happier if the values are checked before getUserObject is called. It seems cleaner to me. Additionally getUserObject should throw an exception if the class could not be loaded. -
Sebastian Michaelsen commented on patch set 3 07/24/12, 17:53
There are several spots like this: t3lib_div::getUserObj($GLOBALS['TYPO3_CONF_VARS']['SC_OPTIONS']['tce']['formevals'][$evalData] . ':&' . $evalData); We should at least check the array value with array_key_exists() or isset() before reading its value. -
Sebastian Michaelsen commented on patch set 5 07/24/12, 17:53
Verified; Looks good to me, but someone else must approve @Helmut: Agreed -
Anja Leichsenring uploaded patch set 2 07/24/12, 17:53
-
Mr. Jenkins commented on patch set 1 07/24/12, 17:53
Build Successful https://ci.typo3.org/job/typo3-v4-core-gerrit/4986/ : SUCCESS -
Mr. Jenkins commented on patch set 2 07/24/12, 17:53
Build Successful https://ci.typo3.org/job/typo3-v4-core-gerrit/4987/ : SUCCESS -
Mr. Jenkins commented on patch set 3 07/24/12, 17:53
Build Successful https://ci.typo3.org/job/typo3-v4-core-gerrit/4990/ : SUCCESS -
Mr. Jenkins commented on patch set 4 07/24/12, 17:53
Fails; I would prefer that you didn't submit this Build Failed https://ci.typo3.org/job/typo3-v4-core-gerrit/4991/ : FAILURE -
Mr. Jenkins commented on patch set 5 07/24/12, 17:53
Build Successful https://ci.typo3.org/job/typo3-v4-core-gerrit/4992/ : SUCCESS -
Wouter Wolters commented on patch set 2 07/24/12, 17:53
Looks good to me, but someone else must approve -
Christian Kuhn merged patch set 5 07/24/12, 17:53