ucscGenomeBrowser/kent: v504 Release

Abstract

The position box now accepts bare codon numbers and codon ranges (separated by an underscore or a hyphen) after gene symbols and NM_/ENST accessions RefSeq transcripts now show a "!" and orange codons where the transcript's codon count differs from the genome's because the transcript and genome sequences differ hgLogin: users can change their recovery email and confirm the new address by email; new minimal Share-a-link links (BLAT, top-right dialog), and a snapshotCleaner utility to remove old snapshots quickLift now lifts alignment tracks: psl and bigPsl, chain and bigChain, and bigMaf and wigMaf. The lifted alignments are drawn in hgTracks, and the hgc details pages, base alignment views and quickLift difference page show the lifted alignment. Ships off by default behind the new hg.conf setting browser.quickLiftAlignments New bigNet track type: a net of alignments stored in a bigBed (bed6+20, with "type bigNet "), so a track hub can carry a real net with its level structure instead of a maf stand-in. Includes the netToBigNet converter, a bigNet help page, trackDb docs, details pages, and quickLift support. Ships off by default behind the hg.conf setting bigNet Track hub track descriptions, custom track documentation and assembly hub genome descriptions are now passed through a new allowlist HTML filter (htmlSanitize) where they are read, keeping only the elements, attributes and style properties that a survey of all 5,390 public hub description pages showed are needed. Ids in the description are prefixed so they cannot collide with the browser's own, and hubCheck warns about anything removed (118 of the 5,390 pages draw a warning) The track color picker () and chromosome alias names () are now on by default, so mirrors and the GBiB get them too; a mirror can still turn either off in hg.conf New trackDb setting denseClick: each item on a dense row gets its own clickable map box with a details-page link and mouseover, so an item can be opened without expanding the track. Works for bed, bigBed, genePred, bigGenePred, psl and bigPsl, and is gated by the hg.conf setting denseClick (off by default). The HPRC pcLAI track goes back to dense with denseClick on Every session load, in every CGI, copied each track collection's hub file and registered the copy as a new row in hgcentral.hubStatus, which nothing removed; 81% of the 3.1 million hubStatus rows on the RR are these dead rows, and each load took a lock that serialized concurrent loads of a shared session. The copy is now made only when the program writing the file asks for it hgTracks performance: RefSeq gene status is fetched with one query per track instead of one database round trip per gene (); coverage-graph drawing no longer allocates a counter per base of the window, which was up to 38% of the render on a Recommended Track Set page (); and the new hg.conf setting mallocTopPad lets the heap grow in larger steps, cutting render time by about 7% at 32 MB At gene-level zoom, the exon mouseover now shows the exon's c. and codon range instead of "Zoom in" Polishing and small improvements to the new BLAT page going out on October 6th Added an opt-in "Keep only last search" checkbox to the new BLAT search form The hgTrackUi page for a track collection gets Hide all / Show all buttons, and "Apply to all visible" is removed The hgTracks long label now says why a track is in density mode: user choice, window too wide, or too many items Improved use of trf/simpleRepeats to avoid runs that never finish on T2T assemblies with new repeat patterns: chunking is now 500 Kb with 10,000-base overlaps instead of 5 Mb, and annotations that cross chunk boundaries are stitched back together so longer annotations stay whole The gcOnFly track now works on some GenArk assemblies The assemblyList table's indexes are now used correctly, so searches on the assemblySearch page run much faster New "hubtools import session" command line option that prints the hubUrl and the hubCheck command New trackDb "detailsScript" scatterPlot type on the hgc page trackDb variable substitution now applies to hub description pages, with a new ${parentTrack} variable Fall back to the RefSeq Historical protein table when an NP_ accession is a deprecated version A CGI parameter with no value, or an empty pair such as "&&", ran into the parameter after it and lost it silently, or aborted the request with "Mangled CGI input string" when it came last; this is what broke the session link in . The query string parsers, the cookie parser, session backups, refreshNamedSessionCustomTracks and the cart's own parser can now skip such pairs, when the new hg.conf setting skipMalformedCgiPairs is on (off by default) A URL with both db=GCF_... and a session load failed with an unknown database error, and a GenArk session loaded this way came up at the default position instead of the saved one; both are fixed GenArk on mirrors: a mirror's hgcentral has no genarkOrg table, which made the GenArk lookup abort; the table is now optional and is included in hgcentral.sql, and the Convert page no longer shows "Genome: (null)" (). The GenArk liftOver accession list is now built as a list rather than a pre-quoted SQL fragment Saved sessions whose files are reached through a symlinked directory are accepted again; 583 saved multi-region sessions on the RR failed with "No BED or BED URL specified" (, , ). Path canonicalization now matches realpath, including a ".." at the root A "type bigBed N" that declared more fields than the file holds drew nothing and broke the details page; the file's own field count is now used (). The old bigBedOnePath fallback is removed, leaving one bigBed load path, and hgc rejects a bigBed type line declaring fewer than three fields The same hgTracks request could return two different images, because a composite subtrack's visibility was written by the main thread while a parallel loader read it; the visibility is now settled before the loaders start quickLift fixes: a second lift now updates tracks already in the quickLift hub, so subtracks turned on later and visibility changes reach the target (); lifting a track inside a container no longer hides everything in it (); hgTrackUi no longer stops partway for a track lifted from a GenArk assembly (); hgc shows the source hub's description again for a track lifted from GenArk (); quickLifted bigWig blocks are placed by the window instead of stretched across it (), and a chain with no aligned block in the window no longer crashes hgTracks (). The quickLift help page now lists the formats that lift Follow-up fixes to the alignment lift: a lifted chain track is colored by normScore like a native one, a reverse-complemented protein alignment keeps its target strand so its blocks draw at full length, a stanza with quickLiftDb but no quickLiftUrl no longer crashes hgTracks or gives a missing-table error on the details page hgc fixes: the RNA fold diagram appears again, and five details-page handlers no longer abort when a chain or net track's other assembly is not a local database, which was 112 of the failures the TrackCheck robot reports (). The knownGene details page no longer reads uninitialized memory, which made an OMIM link appear and disappear between requests Fixes from the v503 Preview II code review: scheme-relative return URLs are refused, uploaded file names are encoded in hgSession and hgGenome, the VCF details page no longer reads past a short tabular INFO value, hubEncode covers custom tracks as well as hub tracks, and the Track Search title is no longer escaped twice Memory-safety fixes: the FASTA reader could grow its buffer too little for a long line (); fifteen calls passed byte sizes instead of element counts to the chop routines, as did one in hubCheck that a hub with many warnings could overflow; the exon mouseover read one element past the end of exonFrames (); and sessionDataSaveTrashFile returned memory that its callers freed with the wrong allocator Updated the in-house RepeatModeler to 2.0.9 and RepeatMasker to 4.2.4; new GenArk browsers will be built with these versions Reordered priorities in the assemblyList table to account for NCBI "reference" status Improved error handling for bad responses from back-end API queries in the liftRequest functions Fixed an illegal reference to a NULL cart in cartJson.c for the API search functions Fix an hgTrackUi 400 error when fetching faceted composite metadata on a curated hub assembly hubCheck no longer resolves bigDataUrl twice, fixing a false "missing file" error for local hubs with relative paths; test added hubCheck now warns about an unreachable bigDataUrl instead of silently skipping it; test added hgBlat: busy spinner during a search, and a shorter PTP4A3 example that shows alt/fix rows The top-right account dialog closes when a link is clicked () and stays inside the blue bar on small screens hgc: the iframeUrl frame now shows at the top of the page and works without a "url" setting The hgCustom assembly search box now always submits the form when an assembly is selected, so you land on the management page if you already have custom tracks on that assembly hubtools now sends the assembly with each upload, so hubSpace rows get a db hubSpace now double-encodes user names when building hub URLs The Sessions page description stayed in the cart, so every session saved afterwards carried it; 41 sessions on hgwdev had another session's description. It is now cleared An explicit "itemRgb on" now wins over a "color" setting on the same track, as it did before 2025 New hg.conf setting "captchaDebug", off by default, to help debug future captcha issues Documented that cdsStart == cdsEnd marks a non-coding transcript in the genePred .as/.h files and the FAQ Improved the gcOnFly description page to fully explain the track's data downloads The trackDb docs listed the wrong track types for about sixty settings; the type lis