allow LIEF to fail when using patchelf

[minrk]

avoids unconditional failure when LIEF raises, even when not used

Description

related to #5175, this allows LIEF to fail just like patchelf.

Previously, even when specifying method=patchelf, LIEF was unconditionally run. This adds error handling matching patchelf, so errors in an unused method don't prevent proceeding with patchelf.

[codspeed-hq]

CodSpeed Performance Report

Merging #5176 will not alter performance

_{Comparing minrk:safe-lief-fail (a538ffd) with main (eca014a)}

Summary

✅ 3 untouched benchmarks

[beeankha]

Hi @minrk , thank you so much for this PR! A few things are requested (or are open for discussion) prior to merge:

1. Is there any way that this code could be re-worked to use log.warn / log.error (with log = utils.get_logger(__name__)) for the error handling? I understand that you most likely wanted to remain consistent with the rest of the mk_relative_linux function, but we can take this opportunity to improve error handling patterns (if you feel inclined to do so!)

2. It would be great to see a test or two for this change; addtionally it might help to have more details in the PR description for steps on how to reproduce LIEF being unconditionally run even when method=patchelf was specified

3. Please add a news file with a brief description of this change

[minrk]

I switched it to log and added news. I'm afraid I don't know how to exercise this failure in a test.

[jaimergp]

Hm, the modified function is now using two different methods to report errors and warnings to the user. This implies slightly different formats, and also different streams (logging will print to stderr and print to stdout, as currently configured). I'd rather stick to one of the methods so:

• Either change everything to logging (and move logger to the top-level module namespace, under the imports)
• Revert to print with the same format as the other messages. We would need to create an issue to tackle a general move to logging, if one doesn't exist yet

[beeankha]

Per Jaime's comment, I reverted the logging changes; I also went ahead and filed an issue to change all error-reporting to logging.

[jaimergp]

Apparently this prints to stderr by default. Just mentioning for awareness. I'm not sure if we should print to stdout for consistency or not.