RFR: 7903765: wget failed in build.sh in jtreg

[Jonathan Gibbons]

On Tue, 2 Jul 2024 13:03:06 GMT, Yasumasa Suenaga <ysuenaga at openjdk.org> wrote:

> I tried to build jtreg, but it failed as following:
> 
> 
> $ bash make/build.sh --jdk /usr/lib/jvm/java-22-ope
> njdk
> [build.sh][INFO] CYGWIN_OR_MSYS=0
> [build.sh][INFO] JAVA_HOME: /usr/lib/jvm/java-22-openjdk
> [build.sh][INFO] Downloading https://archive.apache.org/dist/ant/binaries/apache-ant-1.10.8-bin.zip to /home/ysuenaga/github-forked/jtreg/make/../build/deps/apache-ant-1.10.8-bin.zip
> [build.sh][ERROR] wget exited with exit code 1
> 
> 
> I tried it on Fedora 40, wget is from wget2-wget-2.1.0-9.fc40.x86_64 . Related path seems not to be accepted in this version at least.

Catching up with this thread ...

Overall, while @jaikiran's solution is more complex, I think it is a better and broader solution.

While the original proposal is simpler, it is less obvious -- when I was reading the ibnitial patch (before seeing the whole thread here)  I was wondering why `realpath` was apparently necessary in this specific situation.  So while I agree that `..` is natural and should be OK to use, it is clear that at least one tool does not like the use of `..` for whatever reason it thinks that `..` is bad.  If nothing else, if we deem that `..` is bad, we should at least avoid it at the point at which it is being introduced into paths, and not at the point of use for tools that don't like it.

-------------

PR Comment: https://git.openjdk.org/jtreg/pull/211#issuecomment-2246185110