On Fri, 2006-04-28 at 17:28 -0400, Behdad Esfahbod wrote: > Is there an unwritten invariant that every variable defined in a > configure.in file should survive AC_SUBST? I don't know why you > keep blaming autotools. As a user I don't care what autotools > does or doesn't. What I care is that the new version of intltool > breaks my app.
Your app was already broken. The new version of intltool exposes a bug in your app, which you should fix in your app. And yes, every variable written in a configure.in file should survive being passed to sed. You are not a user who shouldn't care what autotools does or does not do. You are a developer who is using autotools as one of the tools for maintaining a piece of software. If you don't care to know what the tool you are using does or does not do, you shouldn't use it. Would you build a house without knowing how a hammer works? > What you are saying is like we change g_unichar_isalpha in glib > to call libc's isalpha and when people complain that it doesn't > work as expected on some systems, we point them to libc... A bug in libc is a bug in libc, regardless of how your code exposes it. If a change in glib to fix a bug, breaks the code you had to work around the bug, because the data you are passing is bad, doesn't mean glib should let you pass bad data. The bug is in glibc, and I wouldn't call it a regression. Not supporting broken code is not a regresssion. -- dobey _______________________________________________ desktop-devel-list mailing list [email protected] http://mail.gnome.org/mailman/listinfo/desktop-devel-list
