preprocessor: require absolute paths for upload files

When a file is uploaded, it is currently uploaded relative to the
current working directory of a multi-step script. For example, if a
script step changes directory, subsequent uploads (with relative paths)
are processed relative to that new working directory. This is incredibly
confusing and wrong, because it is entirely counter to the expectation
of the script writer who is most likely expecting txtar-like semantics.

Switch to those semantics, that the txtar-specified file is relative to
the starting working directory ($HOME) unless specified as absolute.

This requires a fix in
content/docs/tutorial/validating-simple-yaml-files/en.md which was
relying on the broken previous behaviour.

Preprocessor-No-Write-Cache: true
Signed-off-by: Paul Jolly <[email protected]>
Change-Id: I83d07aec141e52acc08b5ed54671d7a3cf97adbd
Dispatch-Trailer: {"type":"trybot","CL":1176743,"patchset":8,"ref":"refs/changes/43/1176743/8","targetBranch":"alpha"}

content/docs/tutorial/validating-simple-yaml-files/en.md

@@ -65,7 +65,7 @@ Place this information in it, including the deliberate mistake in the `species`
 field:
 
 {{{with upload "en" "yamlBroken"}}}
--- charlie.yml --
+-- validating-yaml-with-cue/charlie.yml --
 name:
   first: Charlie
   last:  Cartwright
@@ -79,7 +79,7 @@ Create a file called `pets.cue` to hold your schema, and place this CUE in it:
 
 {{{with upload "en" "pets.cue"}}}
 #codetab(pets.cue) linenos="table"
--- pets.cue --
+-- validating-yaml-with-cue/pets.cue --
 species!: "cat" | "dog"
 age?:     number
 {{{end}}}
@@ -151,7 +151,7 @@ Update `charlie.yml` to:
 Your corrected data file should read as follows:
 
 {{{with upload "en" "yamlFixed"}}}
--- charlie.yml --
+-- validating-yaml-with-cue/charlie.yml --
 name:
   first: Charlie
   last:  Cartwright
@@ -189,7 +189,7 @@ Add a second data file containing the details of another pet.
 Create a data file named `toby.yml` to hold Toby the dog's details:
 
 {{{with upload "en" "yaml2"}}}
--- toby.yml --
+-- validating-yaml-with-cue/toby.yml --
 name:
   first: Toby
   last: Dog

hugo/content/en/docs/tutorial/validating-simple-yaml-files/index.md

@@ -65,7 +65,7 @@ Create a data file named `charlie.yml` to hold Charlie the cat's details.
 Place this information in it, including the deliberate mistake in the `species`
 field:
 
-```yml { title="charlie.yml" }
+```yml { title="validating-yaml-with-cue/charlie.yml" }
 name:
   first: Charlie
   last:  Cartwright
@@ -77,7 +77,7 @@ age: "15"
 {{< step stepNumber="4" >}}
 Create a file called `pets.cue` to hold your schema, and place this CUE in it:
 
-```cue { title="pets.cue" linenos="table" }
+```cue { title="validating-yaml-with-cue/pets.cue" }
 species!: "cat" | "dog"
 age?:     number
 ```
@@ -158,7 +158,7 @@ Update `charlie.yml` to:
 
 Your corrected data file should read as follows:
 
-```yml { title="charlie.yml" }
+```yml { title="validating-yaml-with-cue/charlie.yml" }
 name:
   first: Charlie
   last:  Cartwright
@@ -195,7 +195,7 @@ Add a second data file containing the details of another pet.
 
 Create a data file named `toby.yml` to hold Toby the dog's details:
 
-```yml { title="toby.yml" }
+```yml { title="validating-yaml-with-cue/toby.yml" }
 name:
   first: Toby
   last: Dog

internal/cmd/preprocessor/cmd/rootfile.go

@@ -24,7 +24,6 @@ import (
 	"math/rand"
 	"os"
 	"os/exec"
-	"path"
 	"path/filepath"
 	"runtime"
 	"slices"
@@ -433,21 +432,40 @@ func (rf *rootFile) buildMultistepScript() (*multiStepScript, error) {
 			f := n.effectiveArchive.Files[0]
 
 			cmdEchoFence := rf.getFence()
+
+			// The upload target path provided via the txtar filename _might_ be
+			// absolute. We don't know.  In case it is, we should treat it as
+			// such. Otherwise, as a convenience we treat it as relative to the
+			// starting working directory of the script, i.e. $HOME. The problem
+			// is that we can only determine if the target is absolute in the script
+			// itself (allowing for env var expansion, etc). So we use a unique
+			// variable name to store the target file path, which is made absolute
+			// as required.
+			targetFileVar := "target" + rf.getFence()
+			pf("if [[ \"%s\" != /* ]]; then %s=$HOME/%s; else %s=%s; fi\n", f.Name, targetFileVar, f.Name, targetFileVar, f.Name)
+
+			// First echo the commands that we will run to make debugging
+			// easier. Notice this block is surrounded in a no-interpolation
+			// cat using a pseudo-random string fence.
 			pf("cat <<'%s'\n", cmdEchoFence)
 			if strings.Contains(f.Name, "/") {
-				pf("$ mkdir -p '%s'\n", path.Dir(f.Name))
+				pf("$ mkdir -p $(dirname $%s)\n", targetFileVar)
 			}
-			pf("$ cat <<EOD > %s\n", f.Name)
+			pf("$ cat <<EOD > $%s\n", targetFileVar)
 			pf("%s\n", f.Data)
 			pf("EOD\n")
 			pf("%s\n", cmdEchoFence)
+
+			// Now write the actual commands to the file.
 			fence := rf.getFence()
 			if strings.Contains(f.Name, "/") {
-				pf("mkdir -p '%s'\n", path.Dir(f.Name))
+				pf("mkdir -p $(dirname $%s)\n", targetFileVar)
 			}
-			pf("cat <<'%s' > %s\n", fence, f.Name)
+			pf("cat <<'%s' > $%s\n", fence, targetFileVar)
 			pf("%s\n", f.Data)
 			pf("%s\n", fence)
+
+			// Check the exist code
 			pf("%s=$?\n", exitCodeVar)
 			pf("if [[ $%s -ne 0 ]]\n", exitCodeVar)
 			pf("then\n")

internal/cmd/preprocessor/cmd/testdata/execute_check_upload_structure.txtar

@@ -0,0 +1,92 @@
+# Ensure that we create uploaded files in the right place,
+# regardless of working directory at the time of the upload.
+
+unquote content/dir/en.md
+
+exec preprocessor execute
+
+cmp hugo/content/en/dir/index.md golden/hugo/content/en/dir/index.md
+
+-- hugo/.keep --
+-- content/site.cue --
+package site
+-- content/dir/page.cue --
+package site
+
+content: dir: page: {
+	leftDelim:  "{{{"
+	rightDelim: "}}}"
+}
+-- content/dir/en.md --
+>---
+>title: JSON Superset
+>---
+>
+>{{{with script "en" "1"}}}
+>find $HOME -type f | sort
+>{{{end}}}
+>
+>{{{with upload "en" "2"}}}
+>-- one.txt --
+>{{{end}}}
+>{{{with upload "en" "3"}}}
+>-- another/two.txt --
+>{{{end}}}
+>
+>{{{with script "en" "4"}}}
+>find $HOME -type f | sort
+>cd another
+>{{{end}}}
+>
+>{{{with upload "en" "5"}}}
+>-- one.txt --
+>{{{end}}}
+>{{{with upload "en" "6"}}}
+>-- another/two.txt --
+>{{{end}}}
+>
+>{{{with script "en" "7"}}}
+>find $HOME -type f | sort
+>{{{end}}}
+>
+-- golden/hugo/content/en/dir/index.md --
+---
+title: JSON Superset
+---
+
+```text { title="TERMINAL" codeToCopy="ZmluZCAkSE9NRSAtdHlwZSBmIHwgc29ydAo=" }
+$ find $HOME -type f | sort
+/home/runner/.bash_logout
+/home/runner/.bashrc
+/home/runner/.profile
+```
+
+```txt { title="one.txt" }
+```
+```txt { title="another/two.txt" }
+```
+
+```text { title="TERMINAL" codeToCopy="ZmluZCAkSE9NRSAtdHlwZSBmIHwgc29ydApjZCBhbm90aGVyCg==" }
+$ find $HOME -type f | sort
+/home/runner/.bash_logout
+/home/runner/.bashrc
+/home/runner/.profile
+/home/runner/another/two.txt
+/home/runner/one.txt
+$ cd another
+```
+
+```txt { title="one.txt" }
+```
+```txt { title="another/two.txt" }
+```
+
+```text { title="TERMINAL" codeToCopy="ZmluZCAkSE9NRSAtdHlwZSBmIHwgc29ydAo=" }
+$ find $HOME -type f | sort
+/home/runner/.bash_logout
+/home/runner/.bashrc
+/home/runner/.profile
+/home/runner/another/two.txt
+/home/runner/one.txt
+```
+

internal/cmd/preprocessor/cmd/testdata/execute_multistagescript.txtar

@@ -220,7 +220,7 @@ package site
 						"in-subdir":        "Zwiy4x77+BFu61goz1QtYAalWU2pz5BnNGntPD/6asM="
 					}
 					multi_step: {
-						"377NPLQLLSANR3SP4BDR88KQU32BUBAPDARQ3K4LHMK8MF2VI6F0====": [{
+						"CCDBDCU5E1MLC2JR58VL9NCT6F49M45C3U8H598PETOOFTBJCR1G====": [{
 							doc: """
 									# script doc comment
 									#scripttag