On 5 May 2011 08:32, Colomban Wendling lists.ban@herbesfolles.org wrote:
Le 04/05/2011 22:53, Enrico Tröger a écrit :
Hi,
I'd say let's make a 0.20.1 bugfix release soon. So we can release the fixes made since the 0.20 release for users.
Have you some particular fixes in mind? Just curious ^^
BTW, I'm not sure it's really important or our very concern, but it seems 0.20 has some weird problems when saving to a Windows machine from a Linux one (or whatever, I haven't investigated the problem), probably due to use of g_file_replace_contents() (which has other advantages...). A few links:
https://sourceforge.net/tracker/?func=detail&aid=3184594&group_id=15... https://sourceforge.net/tracker/?func=detail&aid=3126535&group_id=15...
Yes its in GIO. I remember looking at the source when we were trying to figure out safer file saving without taking forever. in it there is:
#ifdef G_OS_WIN32 close file rename file #else rename file #endif
because windows can't rename a file thats open, the error that both are getting. The test is the *host* os but in the two cases above GIO was built for unix and the file was on windows so the test is wrong so it fails.
Unfortunately i you switch to using safe file saving g_file_set_contents has a similar error because windows can't rename over an existing file, so on windows the old file must be deleted first, but again the decision is based on the host not the target. (you can tell replace_contents or set_contents by the temp file name, replace has "outputstream" in it, set has the original file name)
You have to build Geany without GIO so it uses our unsafe code not replace_contents, I thought there was an option, but I can't find it.
Cheers Lex