[DRAFT] fix(core): support multiple base64 RFCs in Artifact serialization

[greghopkins]

BrowserStack is returning some screenshot data in base64 with the older RFC 2045 style (see the variants summary table), which includes mandatory newlines every 76 characters.

More generally, in this style, whitespace (including newline) is ignored/irrelevant:

$ echo "this is my cool string" | base64
dGhpcyBpcyBteSBjb29sIHN0cmluZwo=

$ echo -e "dGhpcyBpc\\nyBteSBjb29sIH   N0cmluZwo=" | base64 -D
this is my cool string

@jan-molak I'm not quite sure the best way to suggest to handle this... I see some options:

1. Sanitize the input strings by removing all whitespace
2. Modify the regex in looksLikeBase64Encoded() which is what I started to do in this PR
3. Something else?

Thoughts?

[greghopkins]

@jan-molak The main issue with what's proposed here is this assertion:

photo.map(value => expect(value.toString('base64')).to.equal(photo.base64EncodedValue));

base64EncodedValue is a member variable that includes any extra whitespace. value.toString('base64') produces the "normalized" version that includes no whitespace, failing.

[jan-molak]

Hello @greghopkins! Nice catch. Hmm, I think I'd be leaning more towards sanitising the input rather than relaxing the constraint.

This would require us to change:

• WebdriverIO implementation

  serenity-js/packages/webdriverio/src/screenplay/models/WebdriverIOPage.ts

  Line 150 in 06ae5d8

  return await this.browser.takeScreenshot();

• Protractor implementation

  serenity-js/packages/protractor/src/screenplay/models/ProtractorPage.ts

  Line 213 in 06ae5d8

  return promised(this.browser.takeScreenshot());

• and leave Playwright alone since it uses a different protocol (I don't think you can use Playwright with BrowserStack, can you?)

To avoid code duplication we could introduce a protected method in the base Page class like Page.base64Rfc2045ToBase64(...) to sanitise the result of this.browser.takeScreenshot(), or add such converter to core/io instead.

What do you think?