[icedtea-web] RFC: PR804

[Jiri Vanek]

On 10/20/2011 09:00 PM, Omair Majid wrote:
> On 10/18/2011 10:01 PM, Dr Andrew John Hughes wrote:
>> On 18:05 Tue 18 Oct , Omair Majid wrote:
>>> Hi,
>>>
>>> The attached patch fixes PR804, which is a bug in the javaws launcher.
>>> The launcher was eating up spaces in arguments, which caused problems
>>> when javaws was used to run jnlp files which included spaces. The fix
>>> uses bash arrays to hold and build the command to execute to ensure that
>>> spaces are handled correctly.
>>>
>>> Please note that this switches the interpreter used in javaws from sh to
>>> bash.
>>>
>>> I would like to push this to HEAD and backport it to 1.1 as well.
>>>
>>> ChangeLog:
>>> 2011-10-18 Omair Majid<omajid at redhat.com>
>>>
>>> * launcher/javaws.in: Use bash arrays to store and join the various
>>> arguments to ensure spaces are handled correctly.
>>>
>>> Any thoughts or concerns?
>>>
>>> Cheers,
>>> Omair
>>
>> I don't see a huge problem, but I do wonder if a less disruptive
>> solution might be to improve the option parser in IcedTea-Web itself:
>
> That is an interesting approach. I am just afraid of going down this path because it makes icedtea-web different from every other program in terms of how it looks at its arguments.

I agree with Omair, it will be really unpredictable behaviour for everybody else.

>
>>
>> private static String getJNLPFile() {
>>
>> if (args.length == 0) {
>> System.out.println(helpMessage);
>> System.exit(0);
>> } else if (args.length == 1) {
>> return args[args.length - 1];
>> } else {
>> String lastArg = args[args.length - 1];
>> String secondLastArg = args[args.length - 2];
>>
>> if (doubleArgs.indexOf(secondLastArg) == -1) {
>> return lastArg;
>> } else {
>> System.out.println(helpMessage);
>> System.exit(0);
>> }
>> }
>> return null;
>> }
>>
>> (from net.sourceforge.jnlp.runtime.Boot).
>>
>> Instead of assuming that the filename is contained within the last
>> argument, you could step back until you find the last option.
>> The filename is then either the concatenation of all the arguments
>> after that, or all those after that but one if the option is
>> a member of doubleArgs.
>>
>> Another interesting thing about the parser is that it treats
>> unrecognised options as jnlp filenames:
>>
>> $ /usr/lib/icedtea6/bin/javaws -dooldstuff
>> Unable to use Firefox's proxy settings. Using "DIRECT" as proxy type.
>> netx: Invalid jnlp file -dooldstuff
>>
>
> That does sound like a bug. I will take a look at it when I get some free time.
>
>> BTW, these issues with quoting are why we now use Perl for the javac.in
>> in IcedTea and probably why Oracle have binary launchers...
>
> Ah, I wasnt aware of the reason perl was being used for javac.in. I suppose I can rewrite the javaws launcher in perl too, but I would rather stick with bash if possible.

What about  IFS shell variable? It have saved me many times... But also it can be considered as hack...
The only big disadvantage I can see is that when lunched java aplication will call some bash script through runtime exec, then all processes lunched inside this script and using Space as separator, will fail.

J.

>
> Thanks,
> Omair