Repository navigation
Conversation
The phpstan job has been red on every branch since April, and the errors kept changing because composer.phpstan.json asked for ^2.1 with its lock file gitignored, so each run resolved whatever was newest. In September it reported two curl_getinfo stub errors at Client.php:156-157; it now reports two entirely different ones. Pinning to an exact version stops the target moving. Against 2.3.4 the two remaining findings are correct, not false positives: 127 Comparison operation ">=" between 0|1 and 0 is always true 199 Unreachable statement - code above always terminates Every path through the Client::request retry loop returns, continues or throws, and a continue only happens while $retries > 0. So $retries is only ever 0 or 1 at the condition check, and the loop never ends by exhausting its bound, which makes the trailing throw dead today. Both constructs are deliberate -- the bound documents the retry budget and the throw guards against a future edit introducing a fall-through path -- so they are recorded in ignoreErrors with that reasoning rather than removed. Worth a maintainer decision if you would rather drop the throw. Verified: phpstan reports no errors on both PHP 7.4 (what CI uses) and 8.3.
Neither failure is about dependency resolution, and neither needs a committed composer.lock. PHP 5.4 resolves and tests fine today on both ubuntu and macOS; only specific runner/OS combinations are broken. macOS, PHP 5.6/7.4/8.0/8.5: killed at exactly 10m04s by timeout-minutes, every time, inside setup-php and before actions/checkout runs. macOS 5.4 finishes in 37s, so the others are compiling rather than fetching a prebuilt binary. Raised the timeout to 30 minutes for both jobs. Windows, PHP 5.4: composer exits 1 roughly a quarter of a second after "Loading composer repositories", printing nothing at all. A resolution failure would explain itself, so this is a crash. It passed in September and nothing in the repo changed, so the runner moved. composer 2.3 raised its floor to PHP 7.2.5 and `tools: composer:v2` installs the newest 2.x, which makes an incompatible composer the most likely cause, so the 5.x jobs now ask for the 2.2 LTS line explicitly. The Windows change is a best guess that CI can confirm or refute quickly. If 5.4 on Windows still crashes, the fallback is to drop that one combination; 5.4 stays genuinely covered by ubuntu and macOS either way.
Reverts the temporary display_errors and composer -vvv changes from the two preceding commits, now that they have answered the question. composer -vvv showed the job reaching Downloading https://repo.packagist.org/packages.json and the process then exiting 1 with no output whatsoever, 0.17s later. Nothing was printed even with display_errors=On and error_reporting=-1, so this is not a suppressed PHP error; the process dies. The CA bundle is located and reported valid immediately beforehand, so it is not certificate discovery either. PHP 5.4.45 on Windows NT 6.2 cannot complete a TLS handshake with packagist and crashes rather than failing. On ubuntu and macOS, PHP 5.4 links the system OpenSSL and passes. Packagist's TLS requirements are not ours to change and the PHP 5.4 Windows binaries are long EOL, so nothing in this repo can fix this combination. It passed in September; the runner image moved, not the repo. PHP 5.4 support is unaffected. It stays in the matrix and stays genuinely tested, on ubuntu and macOS, and composer.json still declares >=5.3.0.
…ners on: [push, pull_request] fired two identical runs for every push to a PR branch. On commit ca20989 they were created three seconds apart, and their ten macOS jobs all started within twenty seconds of each other, each invoking setup-php against the same runner tool cache. The result was that the same commit produced opposite macOS outcomes: pull_request run 5/5 macOS jobs passed push run 5/5 macOS jobs failed The failures were "composer: command not found" at composer validate -- setup-php reported success but left composer off the PATH, which is what a lost race over the tool cache looks like. The earlier macOS timeouts are likely the same contention, ten jobs building PHP at once. Restricting push to master gives PR branches exactly one run, and a concurrency group cancels superseded runs on rapid pushes. Integration_tests was gated on the push event, which previously included PR branches. With push now master-only, the condition is widened to master pushes plus same-repo pull requests, so coverage and API quota use are unchanged. Fork PRs are excluded because they never receive TINIFY_KEY.
Contributor
Author
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Throwaway draft to test whether resolving dependencies under PHP 8.3 with
composer config platform.php 5.4.45, then switching to PHP 5.4 to run the tests, gets the 5.4/Windows cell working. Will be closed once CI reports.