website compilation fails with `error: unmappable character for encoding ASCII` (#2) · Issues · The Tor Project / Network Health / Metrics / Base

website compilation fails with `error: unmappable character for encoding ASCII`

Compiling the metrics website in our reproducible builds system results in:

    [javac] /var/tmp/build/metrics-web-23437dd19d7a/src/main/java/org/torproject/metrics/web/UpdateNews.java:109: error: unmappable character for encoding ASCII
    [javac]             .replaceAll("??", "&times;")
    [javac]                          ^
    [javac] /var/tmp/build/metrics-web-23437dd19d7a/src/main/java/org/torproject/metrics/web/UpdateNews.java:109: error: unmappable character for encoding ASCII
    [javac]             .replaceAll("??", "&times;")
    [javac]                           ^
    [javac] /var/tmp/build/metrics-web-23437dd19d7a/src/main/java/org/torproject/metrics/web/UpdateNews.java:110: error: unmappable character for encoding ASCII
    [javac]             .replaceAll("??", "&sect;")
    [javac]                          ^
    [javac] /var/tmp/build/metrics-web-23437dd19d7a/src/main/java/org/torproject/metrics/web/UpdateNews.java:110: error: unmappable character for encoding ASCII
    [javac]             .replaceAll("??", "&sect;")
    [javac]                           ^
    [javac] /var/tmp/build/metrics-web-23437dd19d7a/src/main/java/org/torproject/metrics/web/UpdateNews.java:111: error: unmappable character for encoding ASCII

Looking at the respective code we see:

            .replaceAll("×", "&times;")
            .replaceAll("§", "&sect;")
            .replaceAll("–", "&ndash;")

. So, yes, the ASCII complaint makes sense, in particular if you don't have an UTF-8 environment. This looks pretty similar to https://github.com/mimno/Mallet/issues/159. And, yes, settings the encoding to UTF-8 for javac in the build.xml file in the base project does indeed fix the problem.

I am not sure, though, whether we should do that given that the base project is used by all the other metrics projects, too. However, I think setting the javac encoding is the right fix if one has non-ASCII characters in sources files.

/cc @irl