[rfc][icedtea-web] add classapth resolving to rust lunchers (and thus make them work)

[Alex Kashchenko]

Hi Jiri,

Sorry for the delay, see notes inline:

On 12/18/18 2:53 PM, Jiri Vanek wrote:
>
> [...]
>
>>>> configure:
>>>> - introduced configure switch --with-itw-libs which determines if you build portable or distribution
>>>> result. The goal is to achieve, that you can coexists  system itw, and in parallel custom portable
>>>> instance by default, without fear of interference. Still the libraries order can be changed in
>>>> runtime pretty much.

Is it possible to add auto-detection here to default to BUNDLED on 
windows? It would be great if "bash configure" without args continue 
working on windows.

Also currently check_config_files_paths test is failing on windows, I 
suggest adding #[cfg(not(windows))] to it.


>>>> * rust
>>>> - inlcuded laoding of ITW_LIBS in compile time, propagating all jars so they can be reusable. Ketp
>>>> xboothclassapth propagated to, but is use donly in debug output.
>>>> - new module jars_helper is here to try to search for resources based  on ITW_LIBS and locations
>>>> where deps are (we currently have windows and linux dirs named differently). About the members and
>>>> order of LOCAL_PATH, I have many concerns. The current one is most free one.
>>>>    - persisted the error of javafx being on normal classapth. Shouldnt it be on bootclassapth? (see
>>>> nashorn around same lines) and see
>>>> http://mail.openjdk.java.net/pipermail/distro-pkg-dev/2018-November/040492.html)
>>>>    - added method which composes cp/bootcp from freshl located members and from java resources (jdk8
>>>> and down only!)
>>>> - main method made aware of this classpaths and so it it is included in params for
>>>> os.spawn_java_process together with main method and  custom arguments from cmdline and ...
>>>>    ***fanfares*** IT WORKS ***fanfares***
>>>> - os dependent modules got : x ; classpathDelimiter
>>>> - moved  TestLogger to utils - is now shared.

dirs_paths_helpers:

Inside path_to_string I think it is better to use the following:

path.to_str().expect(<...>).to_string()


jars_helper:

1. typo i[m]portant

2. I suggest renaming throwaway variables in pattern matching to "_"

3. in resolve_jar get_libsearch is called multiple times, shouldn't it 
be called only once?

4. in resolve_jar, it is better to use:

match pgmdir.parent()

instead of:

match pgmdir.clone().parent()

and clone it later where needed

5. std::vec::Vec is a part of Prelude - no need for package prefix

6. in get_bootcp_members the same match is repeated 4 times, it is 
better to wrap it into some helper

7. in compose_class_path,

while i != members.len()

should be written as:

for (i, mb) in members.iter().enumerate()


hardcoded_paths:

Contents of itw_libsearch_to_enum should go to FromStr implementation.

>
> [...]
>

-- 
-Alex