Make the `pull-reprint` command work with native PHP by fredrikekelund · Pull Request #3881 · Automattic/studio

fredrikekelund · 2026-06-18T08:22:09Z

Related issues

Fixes STU-1814

How AI was used in this PR

I used Claude to help me with the initial implementation. Then I had to slow down to understand the pull-reprint architecture and the changes that Claude made. I went through a few rounds of iterations by hand and with Claude before reaching the final state.

Proposed Changes

The pull-reprint command comes with more assumptions about the PHP runtime than most other Studio features do. The prime example is how it uses PHP-WASM-based parsing to extract constants from runtime.php.

Studio now has two PHP runtimes, and native PHP is slated to become the new default shortly. This PR makes the pull-reprint command work with native PHP. This implementation does not cover all bases, but it works when pulling a typical WordPress.com site. We'll keep iterating to refine the experience.

Testing Instructions

Run npm run cli:build
Run STUDIO_ENABLE_PULL_REPRINT=1 STUDIO_RUNTIME=native-php node apps/cli/dist/cli/main.mjs pull-reprint
Ensure that you can successfully complete a pull, that the site starts and works

Pre-merge Checklist

Have you checked for TypeScript, React or other console errors?

This commit includes a static reprint.phar build, because the latest HEAD in the reprint repo includes a fix we need (that hasn't been released yet)

fredrikekelund · 2026-06-18T08:24:28Z

-	// For sites imported via `studio pull-reprint`, the pull command
-	// persists the computed start options to start-options.json so the
-	// daemon doesn't need to recompute them (which would spin up PHP
-	// WASM to extract runtime.php constants from the imported site).
-	if ( ! config.useExactMountLayout && config.blueprint?.uri ) {
-		try {
-			const optionsPath = path.join( path.dirname( config.blueprint.uri ), 'start-options.json' );
-			if ( fs.existsSync( optionsPath ) ) {
-				const saved = JSON.parse( fs.readFileSync( optionsPath, 'utf-8' ) );
-				if ( saved.useExactMountLayout ) {
-					config.mountsBeforeInstall = saved.mountsBeforeInstall;
-					config.mounts = saved.mounts;
-					config.wordpressInstallMode = saved.wordpressInstallMode ?? config.wordpressInstallMode;
-					config.useExactMountLayout = true;
-					logToConsole( `Loaded persisted start options from ${ optionsPath } before startup` );
-				}
-			}
-		} catch {
-			// Ignore missing or invalid start options and continue with the provided config.
-		}
-	}


We get this data straight from the config object sent over IPC. No need to re-read start-options.json

fredrikekelund · 2026-06-18T08:26:15Z

 } );

-const wordpressInstallModeSchema = z.enum( [
+const wordpressInstallModeSchema: z.ZodType< WordPressInstallMode > = z.enum( [


We cannot derive a schema from the WordPressInstallMode type, but we can ensure that our manually defined schema matches the given type.

fredrikekelund · 2026-06-18T13:34:21Z

I had to upgrade Reprint to the latest release to make my changes work. This new approach ensures we still ship a fixed Reprint version with every release (i.e. it's not updated on the user's machine), but we also don't have to explicitly change our code to get the latest version. I think that's a good approach

I agree with this approach. It was originally asked by @wojtekn here if we should use a tagged version, but given the fact that Reprint is a known dependency, I think it is safe to get the latest version, and we don't have to change our code to upgrade the version.

epeicher

Thanks for solving this @fredrikekelund! This is now required after merging #3786 because the pull-reprint is now failing. With this solution, we would not need #3922.

I have added an early review, and the changes look good to me. I have just left some questions/comments. Please let me know when you resolve the merging conflicts and I can give some testing.

epeicher · 2026-06-23T09:18:04Z

I agree with this approach. It was originally asked by @wojtekn here if we should use a tagged version, but given the fact that Reprint is a known dependency, I think it is safe to get the latest version, and we don't have to change our code to upgrade the version.

epeicher · 2026-06-23T09:24:14Z

 			}

-			reject( new Error( `PHP command failed (code: ${ code })` ) );
+			reject( new Error( `PHP command (${ args.join( ' ' ) }) failed (code: ${ code })` ) );


In case of an error, is the stderr logged somewhere after these changes?

This change was actually a leftover that I reverted.

What happens to the output of the child process depends on the mode option passed to the runPhpCommand() function

…-php

wpmobilebot · 2026-06-24T10:21:48Z

📊 Performance Test Results

Comparing 1100980 vs trunk

app-size

Metric	trunk	`1100980`	Diff	Change
App Size (Mac)	1311.91 MB	1311.89 MB	0.02 MB	⚪ 0.0%

site-editor

Metric	trunk	`1100980`	Diff	Change
load	1108 ms	1094 ms	14 ms	⚪ 0.0%

site-startup

Metric	trunk	`1100980`	Diff	Change
siteCreation	6494 ms	6513 ms	+19 ms	⚪ 0.0%
siteStartup	6553 ms	6558 ms	+5 ms	⚪ 0.0%

Results are median values from multiple test runs.

Legend: 🟢 Improvement (faster) | 🔴 Regression (slower) | ⚪ No change (<50ms diff)

…-php

fredrikekelund · 2026-06-24T10:42:44Z

The changes in this file are necessary to prevent auto_prepend_file from firing twice for requests that target PHP files directly (like /wp-admin/admin-ajax.php).

If auto_prepend_file fires twice, it could cause classes to be redeclared, for example, which triggers an error.

epeicher

Thanks for working on this @fredrikekelund! I have tested, and the pull-reprint works, but when the site is going to start, I'm getting the following error in the logs:

Error: Error establishing a database connection.

I am testing this by running:

node apps/cli/dist/cli/main.mjs pull-reprint --url <site-url>

This happens consistently when pulling a new site and creating a new site in Studio:

epeicher · 2026-06-24T12:17:20Z

 			);
-			preflight = await runPreflight( studioMetadata, apiUrl, secret, verbose );
+			preflight = await runPreflight(
+				SITE_RUNTIME_NATIVE_PHP,


Should we use a dynamic value for this SITE_RUNTIME_NATIVE_PHP? I'm wondering what happens if a user selects the playground runtime in Studio. Not a blocker, just a question for consideration.

We need to iterate on this in another PR. I'm trying to think of ways that we could tie the pull-reprint state management closer to the regular Studio state management (which is site-based rather than pull-based). That would help address this issue

epeicher · 2026-06-24T12:25:33Z

+			const asset = release.assets.find( ( a ) => a.name === 'reprint.phar' );
+			if ( ! asset ) {
+				throw new Error(
+					`No asset found in latest wp-cli-sqlite-command release ${ release.tag_name }`


Wrong message from another Error probably

epeicher · 2026-06-24T17:30:51Z

After some investigation, I have created this draft PR in Reprint that seems to fix the issue described here:

Emit a dispatch-free runtime.prepend.php from the php-builtin applier WordPress/reprint#278

To test it, what I do is:

On that branch, run composer build:phar
Copy reprint.phar to the Studio repo wp-files/reprint/
Run npm run cli:build on Studio
Test by running node apps/cli/dist/cli/main.mjs pull-reprint --url mypurewebsite5.wpcomstaging.com

Hope that helps

fredrikekelund · 2026-06-25T07:39:35Z

Error: Error establishing a database connection.

I could not reproduce this exact error, but the root cause of this was a late change on my part from passing --runtime=nginx-fpm to --runtime=php-builtin. I tested this, but clearly not thoroughly enough.

Your PR seems worth considering, but I changed the option back to --runtime=nginx-fpm for now so that we can land this PR and proceed with the other items on our list.

epeicher

Thanks @fredrikekelund for addressing the comments. This is working perfectly now, I have tested a new pull and an incremental pull and everything works as expected. LGTM!

epeicher · 2026-06-25T08:17:26Z

 			);
-			preflight = await runPreflight( studioMetadata, apiUrl, secret, verbose );
+			preflight = await runPreflight(
+				SITE_RUNTIME_NATIVE_PHP,


epeicher · 2026-06-25T08:31:27Z

Your PR seems worth considering, but I changed the option back to --runtime=nginx-fpm for now so that we can land this PR and proceed with the other items on our list.

I think that PR is not required eventually, your changes fix the original issue, I have closed it and linked back here for the reasoning.

fredrikekelund added 2 commits June 17, 2026 16:54

First working implementation

cdb95d2

This commit includes a static reprint.phar build, because the latest HEAD in the reprint repo includes a fix we need (that hasn't been released yet)

Tweaks

cd6e15b

fredrikekelund requested a review from epeicher June 18, 2026 08:22

fredrikekelund self-assigned this Jun 18, 2026

fredrikekelund commented Jun 18, 2026

View reviewed changes

Comment thread wp-files/reprint/reprint.phar Outdated

fredrikekelund commented Jun 18, 2026

View reviewed changes

Auto-update reprint again

14b63ec

fredrikekelund marked this pull request as ready for review June 18, 2026 13:32

fredrikekelund commented Jun 18, 2026

View reviewed changes

epeicher reviewed Jun 23, 2026

View reviewed changes

epeicher mentioned this pull request Jun 23, 2026

Set Playground runtime on reprint-pulled sites to fix startup DB error #3922

Closed

1 task

fredrikekelund added 2 commits June 24, 2026 11:30

Merge branch 'trunk' into stu-1814-make-pull-reprint-work-with-native…

87d5343

…-php

WIP: Post-merge fixes

ed8be7c

fredrikekelund and others added 3 commits June 24, 2026 12:34

Tweaks

9e5c98d

Fix double auto-prepend-file issue

d09f2bb

Merge branch 'trunk' into stu-1814-make-pull-reprint-work-with-native…

6ce4b29

…-php

fredrikekelund commented Jun 24, 2026

View reviewed changes

fredrikekelund requested a review from epeicher June 24, 2026 11:56

epeicher reviewed Jun 24, 2026

View reviewed changes

fredrikekelund added 2 commits June 24, 2026 21:38

Nit

ab8bf7b

Tweaks

1100980

fredrikekelund requested a review from epeicher June 25, 2026 07:39

epeicher approved these changes Jun 25, 2026

View reviewed changes

fredrikekelund merged commit de57937 into trunk Jun 25, 2026
13 checks passed

fredrikekelund deleted the stu-1814-make-pull-reprint-work-with-native-php branch June 25, 2026 08:43

Uh oh!

Conversation

fredrikekelund commented Jun 18, 2026

Related issues

How AI was used in this PR

Proposed Changes

Testing Instructions

Pre-merge Checklist

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

epeicher left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

wpmobilebot commented Jun 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

📊 Performance Test Results

app-size

site-editor

site-startup

Uh oh!

Choose a reason for hiding this comment

Uh oh!

epeicher left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

epeicher commented Jun 24, 2026

Uh oh!

fredrikekelund commented Jun 25, 2026

Uh oh!

epeicher left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

epeicher commented Jun 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

wpmobilebot commented Jun 24, 2026 •

edited

Loading

epeicher left a comment •

edited

Loading

epeicher commented Jun 25, 2026 •

edited

Loading