* Andrew Haley: > On 5/11/20 8:58 AM, Florian Weimer wrote: >> * Severin Gehwolf: >> >>> Thanks for the review! Yes, generated-configure.sh changes are due to >>> version skew of autoconf being used. I'll try to generate configure on >>> an older machine so as to avoid this before pushing. Does that sound >>> ok? >> >> THe runstatedir changes come and go (even in the jdk8u-dev history), >> depending on whether Debian's autoconf is used, which patches its >> autoconf 2.69 build to handle runstatedir. There hasn't been an >> autoconf upstream release in many, many years, so going back to older >> versions will not solve this. If you want to avoid this change, you >> either have to use Debian's autoconf, or back it out manually. > > I guess it doesn't matter. Having said that, it's a timely reminder > of how random distro-isms end up in OpenJDK by accident.
The flip side is that distributions carry different patches or use sed to patch configure scripts. > And I am *so* *happy* that autoconf is stable upstream. I remember when > it was changing rapidly, and that was a nightmare. A new upstream release is in the works: <https://lists.gnu.org/archive/html/autoconf/2020-03/msg00003.html> It's true that in the past, you had to have multiple autoconf versions installed and use the right version depending on the project. But at least distributions where providing you with a reasonable choice of autoconf (and automake) versions. Today, you can't easily get Debian's autoconf on Fedora and vice versa, which is clearly not an ideal situation. Thanks, Florian
