`setenv("LANG", ...)` too late

[Floessie]

Alberto extended the MultiLangMgr to set the "LANG" environment variable in #4278. This is definitely the way to go, but doesn't fix the problem for me:

Version: 5.3-514-gac1238e7
Branch: dev
Commit: ac1238e7
Commit date: 2018-01-22
Compiler: cc 6.3.0
Processor: x86_64
System: Linux
Bit depth: 64 bits
Gtkmm: V3.22.0
Lensfun: V0.3.2.0
Build type: release
Build flags:  -std=c++11 -march=native -Werror=unused-label -fopenmp -Werror=unknown-pragmas -Wall -Wno-unused-result -Wno-deprecated-declarations -O3 -DNDEBUG
Link flags:  -march=native
OpenMP support: ON
MMAP support: ON

If I place the setenv() in front of gtk_init() it's effective, but not after it.

Best,
Flössie

[heckflosse]

@Floessie Great you found a solution. Why don't you just push it?

[Floessie]

@heckflosse No, that's no real solution, thus I didn't push it. The problem is, that MultiLangMgr depends on the options being ready, while the options depend on things that happen after gtk_init(). Just taking the options loading from init_rt() to the line before gtk_init() made it crash and I didn't have the time to dig deeper yesternight. I'll give it another try this week...

[heckflosse]

@Floessie Did you test whether it works using putenv() instead of setenv()?
I had to use putenv() for windows builds, because setenv() is not available on msys2.

[Floessie]

@heckflosse I'll give it a try! 👍

[Beep6581]

Could be related:
https://discuss.pixls.us/t/rawtherapee-5-3-524-and-gimp-2-9-9-on-ubuntu-studio-17-10/6546

[Floessie]

@heckflosse I've tried it without success (as I already expected). I hope the msys2 supplied putenv() is implemented better than what POSIX mandates.

Neither setenv() nor putenv() are thread-safe, so they have to be called as early as possible.

[heckflosse]

@Floessie it's putenv(const char *_EnvString)

[Floessie]

@heckflosse And the ownership? Always copied?

[heckflosse]

@Floessie don't know. Can't find the source. Just the header file :(

[Floessie]

I found a solution. 🎉

diff --git a/rtgui/main.cc b/rtgui/main.cc
index bd8f381c4..ca05fdef0 100644
--- a/rtgui/main.cc
+++ b/rtgui/main.cc
@@ -218,13 +218,6 @@ int processLineParams ( int argc, char **argv )
 
 bool init_rt()
 {
-    try {
-        Options::load();
-    } catch (Options::Error &e) {
-        std::cout << "ERROR: " << e.get_msg() << std::endl;
-        return false;
-    }
-
     extProgStore->init();
     SoundManager::init();
 
@@ -487,9 +480,6 @@ int main (int argc, char **argv)
     argv2 = "";
 
     Glib::init();  // called by Gtk::Main, but this may be important for thread handling, so we call it ourselves now
-    gdk_threads_set_lock_functions (G_CALLBACK (myGdkLockEnter), (G_CALLBACK (myGdkLockLeave)));
-    gdk_threads_init();
-    gtk_init (&argc, &argv);  // use the "--g-fatal-warnings" command line flag to make warnings fatal
     Gio::init ();
 
 
@@ -649,6 +639,17 @@ int main (int argc, char **argv)
 
     int ret = 0;
 
+    try {
+        Options::load();
+    } catch (Options::Error &e) {
+        std::cout << "ERROR: " << e.get_msg() << std::endl;
+        return false;
+    }
+
+    gdk_threads_set_lock_functions (G_CALLBACK (myGdkLockEnter), (G_CALLBACK (myGdkLockLeave)));
+    gdk_threads_init();
+    gtk_init (&argc, &argv);  // use the "--g-fatal-warnings" command line flag to make warnings fatal
+
     if (remote) {
         char *app_argv[2] = { const_cast<char *> (argv0.c_str()) };
         int app_argc = 1;

Still, the suffix problem should be fixed in multilangmgr.cc.

HTH,
Flössie

[Beep6581]

This sounds like something we should have in 5.4, no?

[Floessie]

Uhm, the tought of having something so big (GTK init shifted) so late troubles me. But it's your decision.

We also need another patch for multilangmgr.cc that first reads LANG, remembers the suffix, and adds it back when setting LANG.

[heckflosse]

@Floessie

We also need another patch for multilangmgr.cc that first reads LANG, remembers the suffix, and adds it back when setting LANG.

diff --git a/rtgui/multilangmgr.cc b/rtgui/multilangmgr.cc
index b4896d69..a760de2e 100644
--- a/rtgui/multilangmgr.cc
+++ b/rtgui/multilangmgr.cc
@@ -115,6 +115,16 @@ const LocaleToLang localeToLang;
 void setGtkLanguage(const Glib::ustring &language)
 {
     auto l = localeToLang.getLocale(language);
+
+    if (l != "C") {
+        Glib::ustring language(getenv("LANG"));
+        if(!language.empty()) {
+            auto suffixPos = language.find_first_of(".");
+            if(suffixPos != Glib::ustring::npos) {
+                l += language.substr(suffixPos, Glib::ustring::npos);
+            }
+        }
+    }
 #ifdef WIN32
     putenv(("LANG=" + l).c_str());
 #else

[Floessie]

Exactly! 👍 I had no idea, if there are oddities with msys2, like the ones for putenv().