File Changes for markd

v502_base to v503_preview (2026-08-10 to 2026-08-17) v503

• src/hg/hgTracks/hgTracks.c
  • lines changed 6, context: html, text, full: html, text
    5440b4ccf5796778b6a65186bf83c17970f00d7a Sun Aug 16 18:43:46 2026 -0700
    Support PDF output in hgRenderTracks via hgt.psOutput. refs #38118 hgRenderTracks had PDF output wired up behind a compile-time PDF_OUTPUT define that was set to 0, so it could only be reached by editing and rebuilding. Select it at runtime with hgt.psOutput=on instead, which is the same variable the hgTracks View->PDF menu item already uses. PNG remains the default. Also return from handlePostscript after doTrackForm when trackImgOnly is set. In that case makeActiveImage has already converted the eps, written the PDF to stdout and unlinked the eps, so the convertEpsToPdf that follows aborted on the missing file. The client never saw it because the PDF was already flushed and hPrintDisable swallowed the error page, but every request logged "Couldn't open ... .eps" to the apache error log. Document the new parameter in the hgRenderTracks section of customTrackText.html and in the hgRenderTracks tip in allTipsRaw.html.
• src/hg/hgTracks/renderMain.c
  • lines changed 6, context: html, text, full: html, text
    5440b4ccf5796778b6a65186bf83c17970f00d7a Sun Aug 16 18:43:46 2026 -0700
    Support PDF output in hgRenderTracks via hgt.psOutput. refs #38118 hgRenderTracks had PDF output wired up behind a compile-time PDF_OUTPUT define that was set to 0, so it could only be reached by editing and rebuilding. Select it at runtime with hgt.psOutput=on instead, which is the same variable the hgTracks View->PDF menu item already uses. PNG remains the default. Also return from handlePostscript after doTrackForm when trackImgOnly is set. In that case makeActiveImage has already converted the eps, written the PDF to stdout and unlinked the eps, so the convertEpsToPdf that follows aborted on the missing file. The client never saw it because the PDF was already flushed and hPrintDisable swallowed the error page, but every request logged "Couldn't open ... .eps" to the apache error log. Document the new parameter in the hgRenderTracks section of customTrackText.html and in the hgRenderTracks tip in allTipsRaw.html.
• src/hg/htdocs/allTipsRaw.html
  • lines changed 1, context: html, text, full: html, text
    5440b4ccf5796778b6a65186bf83c17970f00d7a Sun Aug 16 18:43:46 2026 -0700
    Support PDF output in hgRenderTracks via hgt.psOutput. refs #38118 hgRenderTracks had PDF output wired up behind a compile-time PDF_OUTPUT define that was set to 0, so it could only be reached by editing and rebuilding. Select it at runtime with hgt.psOutput=on instead, which is the same variable the hgTracks View->PDF menu item already uses. PNG remains the default. Also return from handlePostscript after doTrackForm when trackImgOnly is set. In that case makeActiveImage has already converted the eps, written the PDF to stdout and unlinked the eps, so the convertEpsToPdf that follows aborted on the missing file. The client never saw it because the PDF was already flushed and hPrintDisable swallowed the error page, but every request logged "Couldn't open ... .eps" to the apache error log. Document the new parameter in the hgRenderTracks section of customTrackText.html and in the hgRenderTracks tip in allTipsRaw.html.
• src/hg/htdocs/goldenPath/help/customTrackText.html
  • lines changed 5, context: html, text, full: html, text
    5440b4ccf5796778b6a65186bf83c17970f00d7a Sun Aug 16 18:43:46 2026 -0700
    Support PDF output in hgRenderTracks via hgt.psOutput. refs #38118 hgRenderTracks had PDF output wired up behind a compile-time PDF_OUTPUT define that was set to 0, so it could only be reached by editing and rebuilding. Select it at runtime with hgt.psOutput=on instead, which is the same variable the hgTracks View->PDF menu item already uses. PNG remains the default. Also return from handlePostscript after doTrackForm when trackImgOnly is set. In that case makeActiveImage has already converted the eps, written the PDF to stdout and unlinked the eps, so the convertEpsToPdf that follows aborted on the missing file. The client never saw it because the PDF was already flushed and hPrintDisable swallowed the error page, but every request logged "Couldn't open ... .eps" to the apache error log. Document the new parameter in the hgRenderTracks section of customTrackText.html and in the hgRenderTracks tip in allTipsRaw.html.
• src/hg/lib/hui.c
  • lines changed 19, context: html, text, full: html, text
    38bab7f62d250db395c83bd4b6b0d6c08799ef59 Sat Aug 15 20:01:54 2026 -0700
    Make maf UI species arrays dynamic to fix crash on large mafs. refs #38039 The Codon Translation dropdown in wigMafCfgUi built its option list in a fixed nodeNames[512] stack array, so a maf with more than 511 species wrote past the end of the frame and crashed hgTrackUi. Size it from slCount of the species list instead. Three more fixed arrays in the same path silently dropped data rather than crashing: wigMafGetSpecies chopped speciesOrder into species[2000] and speciesGroups into a 1000 entry array, and wigMafSpeciesTable chopped speciesDefaultOff into words[2000]. All now count with chopLineLen and allocate. Verified by driving wigMafCfgUi directly: 700 species segfaulted before and renders now, and 3000 species previously lost everything past 2000. hgTrackUi output for hg38 multiz470way is unchanged.
• src/lib/udc.c
  • lines changed 77, context: html, text, full: html, text
    954265a3a58f3a3a18475faf6106b4b9d485c696 Sun Aug 16 21:47:53 2026 -0700
    udc: resolve '.' and '..' in a remote URL instead of rejecting them. refs #38120 The #38056 cache-escape fix rejected any ".." path component when udc turns a remote URL into a cache path. That also broke hubs doing something legal: a relative trackDb or bigDataUrl that reaches up a level so two assemblies can share a file. The NHGRI T2T hub hit it four times in one afternoon on dev with TMP.MAT/../HG002v1.1/rep.trackDb.txt. Resolve the dot components the way the remote server resolves them and abort only when ".." would climb above the host. The escape #38056 closed stays closed and keeps the same error message, and the cache path now names the resource actually fetched, so a/x/../b and a/b share one cache entry. simplifyPathToDir in portable.h was not reusable here: it expands ~, eats //, strips a trailing /, and leaves /.. alone rather than catching the climb. Cache paths for URLs with no dot component are byte for byte unchanged, which matters because a change there would invalidate every cached file on every node. Verified by deriving cache paths for 16 URLs across the change; the only lines that differ are a/x/../b/file.txt and a/b/.. which used to abort, and a/./b/file.txt which used to cache under a path containing the dot. That last one orphans any existing cache entry for a URL with a "." component, costing one re-download; such URLs are rare and the new path is the correct one.