fix compiler warnings on EL6,EL7,EL8 #76
Conversation
| @@ -42,7 +42,7 @@ class AndingFilter : public Filter | |||
| Filter *stealLastSubfiler(); | |||
| bool accepts(void *data); | |||
| void combineFilters(int count, int andor); | |||
| unsigned numFilters() { return _subfilters.size(); } | |||
| int numFilters() { return _subfilters.size(); } // satisfy -Wsign-compare | |||
There was a problem hiding this comment.
But we cannot have a negative number of filters. I think we want unsigned here. On what line does it complain about the comparison?
| @@ -52,15 +52,15 @@ extern char *log_file; | |||
|
|
|||
| int num_cached_log_messages = 0; | |||
|
|
|||
| LogCache::LogCache(unsigned long max_cached_messages) | |||
| LogCache::LogCache(long max_cached_messages) // satisfy -Wsign-compare | |||
There was a problem hiding this comment.
Unsigned seems correct to me, should look at the comparison instead
| : _max_cached_messages(max_cached_messages) | ||
| , _num_at_last_check(0) | ||
| { | ||
| pthread_mutex_init(&_lock, 0); | ||
| updateLogfileIndex(); | ||
| } | ||
|
|
||
| void LogCache::setMaxCachedMessages(unsigned long m) | ||
| void LogCache::setMaxCachedMessages(long m) // satisfy -Wsign-compare |
There was a problem hiding this comment.
Unsigned seems correct to me, should look at the comparison instead
| @@ -173,7 +173,7 @@ void LogCache::dumpLogfiles() | |||
| it != _logfiles.end(); | |||
| ++it) | |||
| { | |||
| Logfile *log = it->second; | |||
| // Logfile *log = it->second; // satisfy -Wunused-variable | |||
| @@ -49,7 +49,7 @@ double ServiceSpecialDoubleColumn::getValue(void *data) | |||
| if (is_cmk_passive) { | |||
| host *host = svc->host_ptr; | |||
| servicesmember *svc_member = host->services; | |||
| double check_interval = 1; | |||
| // double check_interval = 1; // satisfy -Wunused-variable | |||
| @@ -51,7 +51,7 @@ int32_t ServicelistStateColumn::getValue(int logictype, servicesmember *mem, Que | |||
| { | |||
| contact *auth_user = query->authUser(); | |||
| int32_t result = 0; | |||
| int lt; | |||
| // int lt; // satisfy -Wunused-variable | |||
| @@ -378,7 +378,7 @@ static gboolean by_one_group(gpointer _name, gpointer _hst, gpointer user_data) | |||
|
|
|||
| void TableHosts::answerQuery(Query *query) | |||
| { | |||
| struct hostbygroup **_hg_tmp_storage = (struct hostbygroup **)&(query->table_tmp_storage); | |||
| // struct hostbygroup **_hg_tmp_storage = (struct hostbygroup **)&(query->table_tmp_storage); // satisfy -Wunused-variable | |||
in underlying resulting assembler code, usually comparison of integers are done by substracting them (and therefore converting them to signed integer before) and check whether result is greater/equal/lower than zero. Looks like compilers get more strict here to prevent any accidents by substracting integers which have value >= 2^(bitlength-1), because then the bit for the +/- get lost. You can avoid the compiler warning by casting during comparison, but this is also not a nice way and can cause issues for same reason. |
|
What I am saying is: We should use consistent types throughout the source, and it makes no sense to change an value which is intrinsically non-negative to a signed integer, just because somewhere we are doing a comparison with a signed integer. Instead we should ensure the other variable we are comparing with is also unsigned. For example rather than changing _max_cached_messages to |
feel free to fix the other side...I have only tried to fix where compiler warned me -> cherry pick the other valid changes and adjust the others... |
split-out of #71