RascalLibraryURIResolver prints on System.err and System.out, asyncronously due to lazy class loading and lazy static initializers. This is what breaks assumptions inside the Evaluator (where System.err and System.out have been wrapped).

[jurgenvinju]

rascal>import lang::rascal::tutor::Indexer;
☐ Parsed ExtractInfo.rsc                                                                                       🕘 0:00:00.386 
☑ Generating $GENERATED_PARSER$41854358                                                                        🕛 0:00:00.436 
☑ Generating $GENERATED_PARSER$1920113068                                                                      🕛 0:00:00.435 
☑ Generating $GENERATED_PARSER$1151540380                                                                      🕛 0:00:00.549 
☑ Generating $GENERATED_PARSER$741664820                                                                       🕛 0:00:00.005 
☑ Loading modules                                                                                              🕔 0:00:08.687 
ok
rascal>

The Parsed ExtractInfo.rsc bar is not supposed to be there. It should be one step of Loading modules.

[jurgenvinju]

rascal>import util::Reflective;
☐ Loading modules                                                                                              🕑 0:00:00.000 
☑ Loading modules                                                                                              🕗 0:00:01.849 

The first line is actually a System.err.println from RascalLibraryURIResolver "[INFO] blablabla" which was not expected at that time and place. The bar accidentally overwrites the [INFO] line, but since the output was not observed by the wrapper, the line offsets is off by one now and the latest version of the progress bar is printed one line below.

[jurgenvinju]

1. RascalLibraryURIResolver is printing in System.err as a result of static initialization blocks
2. This can happen even after printing the prompt and executing a command
3. The output is mixed asynchronously with other output (such as progress bars)
4. The output breaks invariants of the TerminalProgressBarMonitor, leading to duplicate printing of bars and/or loss of output.

[jurgenvinju]

In debug mode there are many System.err printlns in the interpreter. These also break the terminal output and progress monitors invariants.

[jurgenvinju]

Still it the printing on System.err by RascalLibraryURIResolver is questionable. We'd do well by removing this scheme altogether via #1916 and then this problem is fixed as well.

[DavyLandman]

I think we should indeed avoid any direct writes to System.err in our codebase. But then again, where to write to? as there can be more than 1 evaluator active per jvm.

[jurgenvinju]

Yes. This code is executed during the static initialization of the URIResolverRegistry's plugins so shared between all the evaluator's in a single JVM. Still lazily though, and perhaps on a parallel thread. So there's that. It must go.

[jurgenvinju]

See #1962 for a beginning of a solution.