Change Details

See PHI748. This reports an issue with the path validation UI in Owners. I've anecdotally experienced something similar. See PHI757. Leaving inline comments on the right hand side of files not touched by diff B in an interdiff between A and B raises an error. See PHI774, which discusses planning the pathway on T5953 and likely T3980. These are large tasks, but should be plannable today. See PHI649. The file path typeahead in repositories can be difficult to use when your repository has a lot of long paths which all start with `/path/to/some/actual/directory/with/source/in/it/com/java/oracle/`. See PHI580. An install would like write API access to object relationships (depends on, fixes, etc). This is complicated, but perhaps plannable, and we may be able to put a reasonable facade with partial support into the upstream. See some prior discussion in T13130. This didn't get much traction previously (see T13137) so the action may just be to punt it for now. See PHI642. An install would like to give "Disable User" permission to all users so everyone can participate in a large "Battle Royale" event. This is an excellent idea but may justify shoring up some aspects of how user accounts are currently edited. See PHI693. An install notes user confusion over the Differential standalone view incorrectly reporting that changeset navigation shortcuts work. Ideally, at least some of them would actually work, but cause navigation between changeset standalone views. See PHI690. An install would like access to Drydock lease metrics, notably the time a lease spent in queue. See also T13073, but this is more narrowly scoped. See PHI790, which reports an inconsistency in reporting between APCu and APC that possibly points at a more significant issue. See PHI788, which asks for some basic UI improvements to the Drydock operations list page. See T12397 and PHI801. The `ssh-auth` cache currently caches path information too aggressively. See PHI801. `bin/conduit call ...` should support an `--as <user>` flag to make testing and profiling authenticated calls easier. --- // Completeish // See PHI754. Empty files get detected as copies. This is likely a `git` issue but we can write it up. (See T13165.) See PHI751. This looks like a straightforward bug with silent mode and inverse transactions / subeditors. See PHI743. An install reports `project.edit` not working properly when setting member lists with more than 100 members. (This was just a caller mistake with other API calls.) See PHI775, about "More Branches". This is already fixed but should get a followup. See PHI749. The heuristic for `isProbablyJSON()` incorrectly detects `.gitmodule` files which start with `[submodule ...]` as JSON. This can be made less aggressive. See PHI785, which asks for explicit "inbound" and "outbound" configuration in `cluster.mailers` for cases where a provider (e.g., MailGun) is desired for inbound but not configured for outbound. See T12907. This is impacting an export and important to make work for T13076, too. See PHI774. An install would like Spaces support in Phriction and Projects. See PHI765. Some of this ended up in T13166, but improving the push logs to show //which// rule blocked a push (and, ideally, make them queryable by rule) is simpler and a better fit for the actual need. See PHI797. An install missed the last modified time in the head of wiki pages, but this is a little weird and there's no precise timestamp available right now. See PHI766. An install is reporting 18MB chunks written to Harbormaster; the expectation is that logs will be chunked at 1MB. Perhaps a bug when a single 18MB block is written in one go? See also PHI777. (The "fail builds that produce too much output" part of this ended up in T13088.) See PHI766. File chunks are never accessible directly and should not be indexed by Ferret, but we spend a significant amount of time indexing them in `bin/harbormaster rebuild-log`. See PHI766. The `--trace` flag should include absolute timestamps since startup in output so it can be more useful as a proxy-for-profiling on systems where XHProf is not available (which is a reasonably wide range of systems, including production-ish systems). See PHI725, which caught a minor issue with parameter validation in `transaction.search`.

See PHI748. This reports an issue with the path validation UI in Owners. I've anecdotally experienced something similar. See PHI757. Leaving inline comments on the right hand side of files not touched by diff B in an interdiff between A and B raises an error. See PHI774, which discusses planning the pathway on T5953 and likely T3980. These are large tasks, but should be plannable today. See PHI649. The file path typeahead in repositories can be difficult to use when your repository has a lot of long paths which all start with `/path/to/some/actual/directory/with/source/in/it/com/java/oracle/`. See PHI580. An install would like write API access to object relationships (depends on, fixes, etc). This is complicated, but perhaps plannable, and we may be able to put a reasonable facade with partial support into the upstream. See some prior discussion in T13130. This didn't get much traction previously (see T13137) so the action may just be to punt it for now. See PHI642. An install would like to give "Disable User" permission to all users so everyone can participate in a large "Battle Royale" event. This is an excellent idea but may justify shoring up some aspects of how user accounts are currently edited. See PHI693. An install notes user confusion over the Differential standalone view incorrectly reporting that changeset navigation shortcuts work. Ideally, at least some of them would actually work, but cause navigation between changeset standalone views. See PHI690. An install would like access to Drydock lease metrics, notably the time a lease spent in queue. See also T13073, but this is more narrowly scoped. See PHI790, which reports an inconsistency in reporting between APCu and APC that possibly points at a more significant issue. See PHI788, which asks for some basic UI improvements to the Drydock operations list page. See T12397 and PHI801. The `ssh-auth` cache currently caches path information too aggressively. See PHI801. `bin/conduit call ...` should support an `--as <user>` flag to make testing and profiling authenticated calls easier. See PHI805. `bin/storage probe` renders tables larger than 100GB awkwardly/incorrectly. --- // Completeish // See PHI754. Empty files get detected as copies. This is likely a `git` issue but we can write it up. (See T13165.) See PHI751. This looks like a straightforward bug with silent mode and inverse transactions / subeditors. See PHI743. An install reports `project.edit` not working properly when setting member lists with more than 100 members. (This was just a caller mistake with other API calls.) See PHI775, about "More Branches". This is already fixed but should get a followup. See PHI749. The heuristic for `isProbablyJSON()` incorrectly detects `.gitmodule` files which start with `[submodule ...]` as JSON. This can be made less aggressive. See PHI785, which asks for explicit "inbound" and "outbound" configuration in `cluster.mailers` for cases where a provider (e.g., MailGun) is desired for inbound but not configured for outbound. See T12907. This is impacting an export and important to make work for T13076, too. See PHI774. An install would like Spaces support in Phriction and Projects. See PHI765. Some of this ended up in T13166, but improving the push logs to show //which// rule blocked a push (and, ideally, make them queryable by rule) is simpler and a better fit for the actual need. See PHI797. An install missed the last modified time in the head of wiki pages, but this is a little weird and there's no precise timestamp available right now. See PHI766. An install is reporting 18MB chunks written to Harbormaster; the expectation is that logs will be chunked at 1MB. Perhaps a bug when a single 18MB block is written in one go? See also PHI777. (The "fail builds that produce too much output" part of this ended up in T13088.) See PHI766. File chunks are never accessible directly and should not be indexed by Ferret, but we spend a significant amount of time indexing them in `bin/harbormaster rebuild-log`. See PHI766. The `--trace` flag should include absolute timestamps since startup in output so it can be more useful as a proxy-for-profiling on systems where XHProf is not available (which is a reasonably wide range of systems, including production-ish systems). See PHI725, which caught a minor issue with parameter validation in `transaction.search`.

See PHI748. This reports an issue with the path validation UI in Owners. I've anecdotally experienced something similar. See PHI757. Leaving inline comments on the right hand side of files not touched by diff B in an interdiff between A and B raises an error. See PHI774, which discusses planning the pathway on T5953 and likely T3980. These are large tasks, but should be plannable today. See PHI649. The file path typeahead in repositories can be difficult to use when your repository has a lot of long paths which all start with `/path/to/some/actual/directory/with/source/in/it/com/java/oracle/`. See PHI580. An install would like write API access to object relationships (depends on, fixes, etc). This is complicated, but perhaps plannable, and we may be able to put a reasonable facade with partial support into the upstream. See some prior discussion in T13130. This didn't get much traction previously (see T13137) so the action may just be to punt it for now. See PHI642. An install would like to give "Disable User" permission to all users so everyone can participate in a large "Battle Royale" event. This is an excellent idea but may justify shoring up some aspects of how user accounts are currently edited. See PHI693. An install notes user confusion over the Differential standalone view incorrectly reporting that changeset navigation shortcuts work. Ideally, at least some of them would actually work, but cause navigation between changeset standalone views. See PHI690. An install would like access to Drydock lease metrics, notably the time a lease spent in queue. See also T13073, but this is more narrowly scoped. See PHI790, which reports an inconsistency in reporting between APCu and APC that possibly points at a more significant issue. See PHI788, which asks for some basic UI improvements to the Drydock operations list page. See T12397 and PHI801. The `ssh-auth` cache currently caches path information too aggressively. See PHI801. `bin/conduit call ...` should support an `--as <user>` flag to make testing and profiling authenticated calls easier. See PHI805. `bin/storage probe` renders tables larger than 100GB awkwardly/incorrectly. --- // Completeish // See PHI754. Empty files get detected as copies. This is likely a `git` issue but we can write it up. (See T13165.) See PHI751. This looks like a straightforward bug with silent mode and inverse transactions / subeditors. See PHI743. An install reports `project.edit` not working properly when setting member lists with more than 100 members. (This was just a caller mistake with other API calls.) See PHI775, about "More Branches". This is already fixed but should get a followup. See PHI749. The heuristic for `isProbablyJSON()` incorrectly detects `.gitmodule` files which start with `[submodule ...]` as JSON. This can be made less aggressive. See PHI785, which asks for explicit "inbound" and "outbound" configuration in `cluster.mailers` for cases where a provider (e.g., MailGun) is desired for inbound but not configured for outbound. See T12907. This is impacting an export and important to make work for T13076, too. See PHI774. An install would like Spaces support in Phriction and Projects. See PHI765. Some of this ended up in T13166, but improving the push logs to show //which// rule blocked a push (and, ideally, make them queryable by rule) is simpler and a better fit for the actual need. See PHI797. An install missed the last modified time in the head of wiki pages, but this is a little weird and there's no precise timestamp available right now. See PHI766. An install is reporting 18MB chunks written to Harbormaster; the expectation is that logs will be chunked at 1MB. Perhaps a bug when a single 18MB block is written in one go? See also PHI777. (The "fail builds that produce too much output" part of this ended up in T13088.) See PHI766. File chunks are never accessible directly and should not be indexed by Ferret, but we spend a significant amount of time indexing them in `bin/harbormaster rebuild-log`. See PHI766. The `--trace` flag should include absolute timestamps since startup in output so it can be more useful as a proxy-for-profiling on systems where XHProf is not available (which is a reasonably wide range of systems, including production-ish systems). See PHI725, which caught a minor issue with parameter validation in `transaction.search`.