Am 04.07.2017 um 19:16 schrieb Emmanuel Bourg: > Le 4/07/2017 à 17:45, Markus Koschany a écrit : > >> I think it is very important that the encoding is handled consistently >> by all build tools. I wanted to fix this for a while now because too >> many times I had to specify UTF-8 explicitly and the build system tried >> to use ASCII. UTF-8 is the de-facto standard encoding in Debian and >> should be a first class citizen in Debian Java too. Thus said I don't >> want to force this on you. Just give me some hints where I have to look >> in the javahelper package and I experiment with the available options >> myself. > > I'm all for consistency and UTF-8 everywhere, but in this case it brings > absolutely no benefit. This will just result in package updates > reverting the encoding to ISO-8859-1. In the end this will not really > advance the UTF-8 cause.
I understand what you intend to say. Even if UTF-8 was the default, libidw-java would require an override with ISO-8859-1 encoding. However we are talking about the general case. What is more likely that a Java project contains an UTF-8 encoded character or ISO-8859-1 character in 2017? How many Chinese characters are ISO-8859-1 encoded? And by the way the Euro symbol or French special characters are only present in ISO-8859-15. Let's choose UTF-8 as the default and fix packages that don't (but really should) use UTF-8. > This is where the default encoding is specified for jh_build if you want > to modify it: > > https://sources.debian.net/src/javatools/0.61/jh_build/#L54 > https://sources.debian.net/src/javatools/0.61/jh_build/#L74 > >> Lately I haven't discovered any issues in Maven, so the above might be >> true. Though I still don't understand why there is an ASCII fallback in >> Maven and maybe other Java tools. It appears to be completely >> anachronistic. > > Maven doesn't fallback to ASCII, it just uses the default platform > encoding, and on the builders it's ASCII. The day the builders switch to > UTF-8 Maven will also use it by default. Now I'm quite puzzled. The builders set the ASCII encoding? I have seen this issue on my system before and I don't use ASCII encoding. There must be a way to always force Maven to use something else than ASCII. I don't believe this is architecture-dependent. I have the feeling we are going round in circles. This bug is about a "unmappable character" build failure. They regularly occur and I remember that people previously asked for assistance to solve them. Your solution was to set the default encoding in javahelper to ISO8859-1. [1] Then I suggested to use UTF-8 instead because this is obviously the superior encoding. I see your point but I still think the better default is UTF-8. I hope tony will read this and find some better arguments because he is also in favor of UTF-8. :) By the way how are people supposed to override the JH_JAVAC_OPTS variable. I just tried export JH_JAVAC_OPTS="-encoding UTF-8" in libidw-java but this doesn't really seem to work. Inability can't be ruled out though.. I also did remember how I fixed "unmappable character"-bugs before. I used: export JAVA_TOOL_OPTIONS=-Dfile.encoding=iso-8859-1 for javahelper projects. To be honest I didn't expect that much resistance against using UTF-8. [1] https://anonscm.debian.org/cgit/pkg-java/javatools.git/commit/?id=baff5c8
signature.asc
Description: OpenPGP digital signature
