.loadlib fails silently

[moritz]

It appears that the .loadlib fails silently on wrong paths. For example Rakudo's perl6.pir contains the lines

 .loadlib 'perl6_group'
 .loadlib 'perl6_ops'

If I introduce a typo in the name perl6_ops in the second line, I get no error message from the failed load, just from subsequent failures.

This makes diagnosis very hard, and IMHO a failed .loadlib should throw an exception.

[wayland]

I'm subscribing to this ticket so that I'll (hopefully) get cc'ed on any changes. Is there an easier way to do this?

[wayland]

It would also be nice if the error message said where it tried looking for the files, so that I know that my -L options are working correctly :)

[wayland]

The patch I've attached seems to solve the problem. It throws an exception instead of printing a warning. The exception contains the path used to search for the library.

The patch also adds a -S option to Parrot that allows more paths to be added to the search path. I found this option useful in building Rakudo.

[wayland]

allison requested via IRC that the -S option be done as a separate patch/ticket. That's now in https://trac.parrot.org/parrot/ticket/442

[jkeenan]

Replying to moritz:

It appears that the .loadlib fails silently on wrong paths. For example Rakudo's perl6.pir contains the lines {{{ .loadlib 'perl6_group' .loadlib 'perl6_ops' }}} If I introduce a typo in the name perl6_ops in the second line, I get no error message from the failed load, just from subsequent failures.

Moritz,

Does this problem still present itself?

Would the non-S part of wayland's patch be of benefit?

Thank you very much.
kid51

[moritz]

Replying to jkeenan:

Replying to moritz:

It appears that the .loadlib fails silently on wrong paths. For example Rakudo's perl6.pir contains the lines {{{ .loadlib 'perl6_group' .loadlib 'perl6_ops' }}} If I introduce a typo in the name perl6_ops in the second line, I get no error message from the failed load, just from subsequent failures.

Moritz, Does this problem still present itself?

Yes.

Would the non-S part of wayland's patch be of benefit?

If it makes .loadlib complain loudly and fatally (unless caught), yes.

[jkeenan]

Replying to moritz:

If it makes .loadlib complain loudly and fatally (unless caught), yes.

Unfortunately, in the 15 months since Wayland first posted his patch, the files being patched have changed so much that 2 of the 3 do not apply cleanly.

[li11-226:parrot] 524 $ patch -p1 < ~/learn/parrot/patches/parrot_libloading.patch 
patching file src/library.c
Hunk #1 FAILED at 710.
1 out of 1 hunk FAILED -- saving rejects to file src/library.c.rej
patching file src/dynext.c
Hunk #1 FAILED at 615.
1 out of 1 hunk FAILED -- saving rejects to file src/dynext.c.rej
patching file include/parrot/library.h
Hunk #1 succeeded at 85 (offset 11 lines).

Wayland, if you are still around: Could you (or anyone else) take another whack at it?

Thank you very much.
kid51

[plobsing]

I've fixed the problem in r47282.

However, I made it such that only failing '.loadlib' directives throw exceptions. The 'loadlib' opcode interface is unchanged.