Xml.PokeInnerText mangles DOCTYPE definition

[kentcb]

Description

In migrating to FAKE 5, I've switched from XmlPokeInnerText to Xml.PokeInnerText. However, the result is that my XML is now left in an invalid state:

BEFORE:

<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">

AFTER:

<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"[]>

Repro steps

Please provide the steps required to reproduce the problem

1. Update an XML file that has a DOCTYPE definition using Xml.PokeInnerText

Expected behavior

Should act same as old XmlPokeInnerText.

Actual behavior

Breaks the XML.

Known workarounds

Use old API.

[matthid]

Can you provide a minimal XML sample and the PokeInnerText call? I feel like this could be a framework bug as I cannot recall any code changes there. Though the [] look quite suspicious to an empty list being printed so I'd like to reproduce that locally before reporting to the corefx team.

[kentcb]

Here's some example XML (Info.plist):

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
    <key>CFBundleIdentifier</key>
    <string>foo</string>
</dict>
</plist>

Here's an example target:

Target.Create "testing" (fun _ ->
    let plist = srcDir @@ "UI.iOS/Info.plist"
    let bundleIdentifier = "whateva"
    Xml.PokeInnerText plist "plist/dict/key[text()='CFBundleIdentifier']/following-sibling::string" bundleIdentifier
)

And here's the (invalid) output:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"[]>
<plist version="1.0">
  <dict>
    <key>CFBundleIdentifier</key>
    <string>whateva</string>
  </dict>
</plist>

[matthid]

This is kind of an unfortunate side-effect of using XmlDocument:

• https://stackoverflow.com/a/16451790
• https://stackoverflow.com/questions/284394/net-xmldocument-why-doctype-changes-after-save

I changed it in a way that without [] is always preferred but realistically that might break people who expect [], unfortunately it seems to be quite hard to leave it as-is.
So this should be fixed with RC015