You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Right now the number is both wrong (1.5.1 when the latest stable version is 4.5.6) and the wrong format (the PDE expects a revision number like 1434 and NOT semantic versioning like 4.5.6).
You can also come discuss this on this Discord thread for quicker back and forth and unformed thoughts. However, please post any relevant insights in the comments below so the findings are not siloed to Discord.
Warning
AI Disclosure: The report below was generated with AI help. Specifically, Claude Opus 5 for the initial report, with verification by Codex Astra GPT 6 High
I have verified that the report is directionally sound but human replication/investigation is necessary. I added some notes in the body of the issue for additional context or where I disagreed with the machine.
readInt calls Integer.parseInt(reader.readLine().trim()). The response 1.5.1 throws NumberFormatException, stopping the check before it compares the latest build with the installed revision.
The background thread catches and ignores the exception. Clients using this endpoint therefore receive no PDE update prompt while it serves this value.
That change replaced the manually ordered selection with a generated list. Its first entry changed from processing-1310-4.4.10 to processing-1.5.1, without changing the build hook that assumes the first entry is the latest release.
Executing the existing build hook with historical selection data confirms:
Even selecting an older tag with a valid numeric build would incorrectly suppress update prompts for newer installed versions.
Suggested fix
Note
Raph: The suggested fix below continues to use selected.json as the source of truth. I don't think that's the right call. We could get the revision number directly from the release tag on GitHub.
Select the highest build number independently of array order, skip tags without a build-number segment, and fail the build if no valid candidate exists:
exports.onPostBuild=()=>{const{ selectedReleases }=require('./content/download/selected.json');constbuild=selectedReleases.map((tag)=>tag.match(/^processing-(\d+)-/)).filter(Boolean).map((match)=>Number(match[1])).sort((a,b)=>b-a)[0];if(!build){thrownewError('latest.txt: no release with a build number in selected.json');}fs.writeFileSync(path.join(__dirname,'public','download','latest.txt'),String(build));};
This produces 1434 for the current selection, including with the array reversed, and throws when no build-number tags exist.
Two additional validation gaps remain:
The release generator does not exclude prereleases from selectedReleases. Stable-release eligibility should be checked explicitly.
Tl;dr: Looks like Processing users may not have been getting update notifications in the PDE since October last year.
Normally, Processing shows a popup when a newer version is available.
The way the PDE knows a new version is available is by reading https://processing.org/download/latest.txt and comparing the revision number with its own.
Right now the number is both wrong (
1.5.1when the latest stable version is4.5.6) and the wrong format (the PDE expects a revision number like1434and NOT semantic versioning like).4.5.6You can also come discuss this on this Discord thread for quicker back and forth and unformed thoughts. However, please post any relevant insights in the comments below so the findings are not siloed to Discord.
Warning
AI Disclosure: The report below was generated with AI help. Specifically, Claude Opus 5 for the initial report, with verification by Codex Astra GPT 6 High
I have verified that the report is directionally sound but human replication/investigation is necessary. I added some notes in the body of the issue for additional context or where I disagreed with the machine.
Summary
https://processing.org/download/latest.txt currently returns:
It should return the integer build number of the latest stable release, currently
1434for Processing 4.5.6.Impact
The PDE update checker reads this endpoint:
UpdateCheck.java:62defaultsLATEST_URLtohttps://processing.org/download/latest.txt.app/build.gradle.kts:66configures the same URL.readIntcallsInteger.parseInt(reader.readLine().trim()). The response1.5.1throwsNumberFormatException, stopping the check before it compares the latest build with the installed revision.The background thread catches and ignores the exception. Clients using this endpoint therefore receive no PDE update prompt while it serves this value.
Processing 3 uses the same path over HTTP; its current behavior also depends on HTTP serving and redirects.
Note
Raph: This could have been a contributing factor to the decline in downloads we have observed beginning in October 2025 (speculative).
Regression
The source regression was introduced by “Automate selection of latest Processing releases” (
bd09d71a, October 15, 2025), then merged on October 16 (4bc0ef7b). The exact production onset requires deployment records.That change replaced the manually ordered selection with a generated list. Its first entry changed from
processing-1310-4.4.10toprocessing-1.5.1, without changing the build hook that assumes the first entry is the latest release.Executing the existing build hook with historical selection data confirms:
latest.txt13101.5.11.5.1Cause
onPostBuildingatsby-node.jswrites:Two assumptions fail:
fetchReleases.jsgroups releases under numeric major-version keys and usesObject.values(), which enumerates those keys in ascending order.processing-1.5.1has no build-number segment, so splitting it writes the version string as the build number.The current selection is:
Even selecting an older tag with a valid numeric build would incorrectly suppress update prompts for newer installed versions.
Suggested fix
Note
Raph: The suggested fix below continues to use
selected.jsonas the source of truth. I don't think that's the right call. We could get the revision number directly from the release tag on GitHub.Select the highest build number independently of array order, skip tags without a build-number segment, and fail the build if no valid candidate exists:
This produces
1434for the current selection, including with the array reversed, and throws when no build-number tags exist.Two additional validation gaps remain:
selectedReleases. Stable-release eligibility should be checked explicitly.processing-13010-4.4.10(adb2c6ea), corrected five minutes later (cbe4958e), would pass this check. Validation against release metadata would provide stronger protection.The hook runs during builds, not local development. After deploying the fix, eligible clients on older builds should resume receiving update prompts.