58104a98358604975dac0a9350ca91d2d25c501d
max
  Thu Sep 10 05:02:45 2026 -0700
hUserAbort shows its message to the user instead of turning into a 500

hUserAbort() reports an error caused by user input, so the message is written
to be read by the user.  It was only reaching them when a CGI had already
pushed a warn handler of its own.  The apiKey and bot checks call it from
main() before that happens, and the default handler then writes to stderr and
nothing else, unless hg.conf sets showEarlyErrors - off by default, and off on
the RR.  Apache turns the empty response into a 500, which is what
hubApi/hubApi.c works around by pre-validating the apiKey itself.

hVaUserAbort() now turns doContentType on for the rest of the process when it
is running as a CGI, so the default handler emits the Content-Type line and
the message.  It stays off for a program that never called cgiSpoof(), and it
is inert inside an errCatch, which pushes its own warn handler - so a caller
that catches the abort to write its own response (hubApi's JSON) is unchanged.

Fixes the va_list handling in defaultVaWarn() while in there.  It read args
three times but only the second and third read from a va_copy: the first
vfprintf consumed args itself, so the two reads after it saw a spent va_list
and the copy sent to the browser lost every %s and %d.  It printed
"Bad thing: [br]" where the message was "Bad thing: %s<br>".  Every read now
takes its own copy, and the buffer is filled with vsnprintf rather than
vsprintf.

No XSS: on this path defaultVaWarn replaces < and > with [ and ] across the
whole formatted message, args included.  The other handlers that can report an
hUserAbort - earlyWarningHandler and cartEarlyWarningHandler via
htmlVaEncodeErrorText, htmlVaWarn, webVaWarn - all run the arguments through
vaHtmlDyStringPrintf, which html-encodes & < > / " and '.  No caller passes
user data as the format string.

diff --git src/hg/lib/hCommon.c src/hg/lib/hCommon.c
index 41e865de9e5..8768a57f18b 100644
--- src/hg/lib/hCommon.c
+++ src/hg/lib/hCommon.c
@@ -1,439 +1,453 @@
 /* hCommon.c - routines used by many files in hgap project. */
 
 /* Copyright (C) 2014 The Regents of the University of California 
  * See kent/LICENSE or http://genome.ucsc.edu/license/ for licensing information. */
 
 #include "common.h"
 #include "hCommon.h"
 #include "chromInfo.h"
 #include "portable.h"
 #include "hgConfig.h"
 #include "errAbort.h"
 #include "htmshell.h"
+#include "cheapcgi.h"
 
 
 static char *_hgcName = "../cgi-bin/hgc";	/* Path to click processing program. */
 static char *_hgTracksName = "../cgi-bin/hgTracks"; /* Path back to genome browser. */
 static char *_hgTrackUiName = "../cgi-bin/hgTrackUi"; /* Path to extended ui program. */
 static char *_hgFileUiName = "../cgi-bin/hgFileUi";   /* Path to downloladable files CGI. */
 static char *_hgTextName = "../cgi-bin/hgText"; /* Path back to the text browser. */
 static char *_hgTablesName = "../cgi-bin/hgTables"; /* Path back to the table browser. */
 static char *_hgVaiName = "../cgi-bin/hgVai"; /* Path back to the variant annotation integrator. */
 static char *_hgCustomName = "../cgi-bin/hgCustom"; /* Path back to the custom tracks manager. */
 static char *_hgCollectionName = "../cgi-bin/hgCollection"; /* Path back to the composite builder */
 static char *_hgHubConnectName = "../cgi-bin/hgHubConnect"; /* Path back to the track hub manager. */
 static char *_hgSessionName = "../cgi-bin/hgSession";	/* Path to session manager. */
 static char *_hgPalName = "../cgi-bin/hgPal"; /* Path back to the protein aligner */
 static char *_hgVarAnnogratorName = "../cgi-bin/hgVarAnnogrator"; /* Path to variant annot intgr */
 static char *_hgIntegratorName = "../cgi-bin/hgIntegrator"; /* Path to annotation intgrator */
 static char *_hgGeneName = "../cgi-bin/hgGene"; /* Path to gene details */
 
 char *hgPalName()
 /* Relative URL to click processing program. */
 {
 return _hgPalName;
 }
 
 char *hgcName()
 /* Relative URL to click processing program. */
 {
 return _hgcName;
 }
 
 char *hgTracksName()
 /* Relative URL to browser. */
 {
 return _hgTracksName;
 }
 
 char *hgTrackUiName()
 /* Relative URL to extended track UI. */
 {
 return _hgTrackUiName;
 }
 
 char *hgFileUiName()
 /* Relative URL to downloladable files UI. */
 {
 return _hgFileUiName;
 }
 
 char *hgTextName()
 /* Relative URL to old table browser. */
 {
 return _hgTextName;
 }
 
 char *hgTablesName()
 /* Relative URL to table browser. */
 {
 return _hgTablesName;
 }
 
 char *hgVaiName()
 /* Relative URL to variant annotation integrator. */
 {
 return _hgVaiName;
 }
 
 char *hgCustomName()
 /* Relative URL to custom tracks manager. */
 {
 return _hgCustomName;
 }
 
 char *hgCollectionName()
 /* Relative URL to composite builder. */
 {
 return _hgCollectionName;
 }
 
 char *hgHubConnectName()
 /* Relative URL to track hub manager. */
 {
 return _hgHubConnectName;
 }
 
 char *hgSessionName()
 /* Relative URL to session manager. */
 {
 return _hgSessionName;
 }
 
 char *hgVarAnnogratorName()
 /* Relative URL to variant annotation integrator program. */
 {
 return _hgVarAnnogratorName;
 }
 
 char *hgIntegratorName()
 /* Relative URL to annotation integrator program. */
 {
 return _hgIntegratorName;
 }
 
 char *hgGeneName()
 /* Relative URL to gene details program (hgGene). */
 {
 return _hgGeneName;
 }
 
 char *hgAbsUrl()
 /* absolute URL to current CGI. Needs to be freed. */
 {
 // get the full URL of this hgTracks page, so external page can construct a custom track
 // and link back to us
 char* host = getenv("HTTP_HOST");
 char* reqUrl = getenv("REQUEST_URI");
 char* isHttps = getenv("HTTPS");
 // remove everything after ? in URL
 if (reqUrl) 
     {
     char *e = strchr(reqUrl, '?');
     if (e) *e = 0; 
     }
 else 
 {
     // when called from command line, cannot get argv so using dummy name
     reqUrl = "/cgi-bin/hgTracks";
     host = "genome.ucsc.edu";
     isHttps = "on";
 }
 
 char *prot = NULL;
 if (isHttps && sameWord(isHttps, "on"))
     prot = "https";
 else
     prot = "http";
 
 char *url = needMem(4000);
 safef(url, 4000, "%s://%s%s", prot, host, reqUrl);
 return url;
 }
 
 char *hgAbsUrlCgi(char *cgiName)
 /* Full absolute URL to another CGI, including the protocol part. Needs to be freed. Example argument: "hgTracks" */
 {
 char *url = hgAbsUrl();
 char *lastSlash = strrchr(url, '/');
 if (lastSlash!=NULL)
 {
     lastSlash++;
     *lastSlash = '\0';
 }
 char *newUrl = catTwoStrings(url, cgiName);
 //freeMem(cgiName);
 return newUrl;
 }
 
 static void finishCloneName(char *fragName, char *e, char cloneName[128])
 /* Finish conversion from frag to clone or clone.ver name. */
 {
 int size;
 
 if (e == NULL)
     e = fragName + strlen(fragName);
 size = e - fragName;
 if (size >= 128)
     errAbort("name too long %s\n", fragName);
 memcpy(cloneName, fragName, size);
 cloneName[size] = 0;
 }
 
 void fragToCloneName(char *fragName, char cloneName[128])
 /* Convert fragment name to clone name. */
 {
 char *e = strchr(fragName, '.');
 finishCloneName(fragName, e, cloneName);
 }
 
 void fragToCloneVerName(char *fragName, char cloneVerName[128])
 /* Convert fragment name to clone.version name. */
 {
 char *e = strchr(fragName, '.');
 if (e == NULL)
     errAbort("No . in fragName %s", fragName);
 e = strchr(e, '_');
 finishCloneName(fragName, e, cloneVerName);
 }
 
 void recNameToFileName(char *dir, char *recName, char *fileName, char *suffix)
 /* Convert UCSC style fragment name to name of file for a clone. */
 {
 char *e;
 char *d = fileName;
 int size;
 
 
 /* Start file name with directory if any. */
 if (dir != NULL)
     {
     size = strlen(dir);
     memcpy(d, dir, size);
     d += size;
     if (dir[size-1] != '/')
 	*d++ = '/';
     }
 if (*recName == '>')
     ++recName;
 recName = skipLeadingSpaces(recName);
 e = strchr(recName, '.');
 if (e == NULL)
     e = skipToSpaces(recName);
 if (e == NULL)
     e = recName + strlen(recName);
 size = e - recName;
 memcpy(d, recName, size);
 d += size;
 strcpy(d, suffix);
 }
 
 void faRecNameToQacFileName(char *dir, char *recName, char *fileName)
 /* Convert fa record name to file name. */
 {
 recNameToFileName(dir, recName, fileName, ".qac");
 }
 
 void faRecNameToFaFileName(char *dir, char *recName, char *fileName)
 /* Convert fa record name to file name. */
 {
 recNameToFileName(dir, recName, fileName, ".fa");
 }
 
 void gsToUcsc(char *gsName, char *ucscName)
 /* Convert from
  *    AC020585.5~1.2 Fragment 2 of 29 (AC020585.5:1..1195)
  * to
  *    AC020585.5_1_2
  */
 {
 char *s, *e, *d;
 int size;
 
 /* Copy in accession and version. */
 d = ucscName;
 s = gsName;
 e = strchr(s, '~');
 if (e == NULL)
     errAbort("Expecting . in %s", gsName);
 size = e - s;
 memcpy(d, s, size);
 d += size;
 
 /* Skip over tilde and replace it with _ */
 s = e+1;
 *d++ = '_';
 
 e = skipToSpaces(s);
 if (e == NULL)
     e = s + strlen(s);
 size = e - s;
 memcpy(d, s, size);
 d[size] = 0;
 subChar(d, '.', '_');
 return;
 }
 
 char *skipChr(char *s)
 /* Skip leading 'chr' in string (to get the actual chromosome part). */
 {
 if (startsWith("chr", s))
     s += 3;
 else if (startsWith("scaffold_", s))
     s += 9;
 else if (startsWith("Scaffold_", s))
     s += 9;
 return s;
 }
 
 int chromToInt(char *s)
 /* converts a chrom name chrXX into an integer from 1 to 54.
     X = 23 Y = 24 Un = 25 M = 26 random = chr + 26;*/
 {
 char *u;
 int ret = 0;
 char str[64];
 
 if (!startsWith("chr", s))
     {
     return 0;
     }
 s += 3;
 safef(str, sizeof(str), "%s", s);
 u = strchr(str,'_');
 if (u != NULL)
     {
     ret = 26;
     *u = '\0';
     }
 switch (str[0])
     {
     case 'X':
         ret += 23;
         break;
     case 'Y':
         ret += 24;
         break;
     case 'U':
         ret += 25;
         break;
     case 'M':
         ret += 26;
         break;
     default:
         ret += atoi(s);
     }
 return ret;
 }
 
 void hTableStart()
 /* Output a table with solid borders. */
 /* For some reason BORDER=1 does not work in our web.c nested table scheme.
  * So use web.c's trick of using an enclosing table to provide a border.   */
 {
 puts("<!--hTableStart-->" "\n"
      "<TABLE BGCOLOR=\"#"HG_COL_BORDER"\" BORDER=\"0\" CELLSPACING=\"0\" CELLPADDING=\"1\"><TR><TD>");
 puts("<TABLE BORDER=\"1\" BGCOLOR=\"#"HG_COL_INSIDE"\" CELLSPACING=\"0\">");
 }
 
 void hTableEnd()
 /* Close out table started with hTableStart() */
 {
 puts("</TABLE>");
 puts("</TD></TR></TABLE>");
 puts("<!--hTableEnd-->");
 }
 
 static boolean stackDumpDisabled = FALSE;  // prevent accidental recursion or undesired dumps
 static boolean hDumpAbortCalled = FALSE;
 
 static void hDumpStackAbortHandler()
 /* abort handle that prints stack dump then invokes the previous abort
  * handler on the stack. */
 {
 if (stackDumpDisabled)
     {
     stackDumpDisabled = FALSE;
     }
 else
     {
     dumpStack("\nStack dump:");
     }
 hDumpAbortCalled = TRUE;
 popAbortHandler(); // remove us from the stack 
 // continue with next abort handler
 noWarnAbort();
 }
 
 boolean hDumpStackEnabled(void)
 /* is browser.pstack enabled?  */
 {
 return cfgOptionBooleanDefault("browser.dumpStack", FALSE);
 }
 
 void hDumpStackDisallow(void)
 /* prevent any dumping of the stack */
 {
 stackDumpDisabled = TRUE;
 }
 
 void hDumpStackPushAbortHandler(void)
 /* push the stack dump abort handler on the stack if it's enabled.  This should be pushed
  * after the warn handle that will do the actual reporting */
 {
 if (hDumpStackEnabled())
     {
     errAbortDebugnPushPopErr();
     pushAbortHandler(hDumpStackAbortHandler);
     }
 }
 
 void hDumpStackPopAbortHandler()
 /* pop the stack dump abort handler from the stack if it's enabled */
 {
 if (hDumpStackEnabled() && !hDumpAbortCalled)
     popAbortHandler();
 hDumpAbortCalled = FALSE;
 }
 
 void hVaUserAbort(char *format, va_list args)
 /* errAbort when a `user' error is detected.  This is an error that comes
  * from user input. This disables the logging stack dumps. */
 {
 hDumpStackDisallow();
+/* A user error is written for the user to read, so it has to reach the browser.  When we are
+ * called before the CGI has pushed a warn handler of its own - the apiKey and bot checks do
+ * this, from main() - the default handler writes only to stderr unless doContentType is set,
+ * and apache turns that empty response into a 500.  Turn it on so the default handler emits
+ * the Content-Type line and the message (with < and > neutered) to stdout.  This is inert
+ * inside an errCatch, which pushes its own warn handler, so a caller that means to catch the
+ * abort and write its own response (hubApi's JSON) still gets to.
+ *
+ * cgiIsOnWeb() is the right test here even though it is TRUE for a spoofed command-line run
+ * of a CGI: the question is "is this a CGI process", and a spoofed run wants the header too.
+ * A program that never calls cgiSpoof() reads FALSE and keeps writing to stderr only. */
+if (cgiIsOnWeb())
+    errAbortSetDoContentType(TRUE);
 vaErrAbort(format, args);
 }
 
 void hUserAbort(char *format, ...)
 /* errAbort when a `user' error is detected.  This is an error that comes
  * from user input. This disables the logging stack dumps. */
 {
 va_list args;
 va_start(args, format);
 hVaUserAbort(format, args);
 va_end(args);
 }
 
 boolean hAllowAllTables(void)
 /* Return TRUE if hg.conf's hgta.disableAllTables doesn't forbid an 'all tables' menu. */
 {
 return !cfgOptionBooleanDefault("hgta.disableAllTables", FALSE);
 }
 
 void cspWriteResponseHeader(void)
 /* Write the Content Security Policy as an http response header, if hg.conf
  * turns it on.  Must be called before the blank line that ends the http header
  * block.  Only the first call in a process writes anything.
  *
  * This exists so that pages which build their own http header block, and so
  * never reach the library code that writes the meta tag, still carry a policy.
  * The gate is off by default so it can be turned on one machine at a time. */
 {
 static boolean written = FALSE;
 if (written)
     return;
 if (!cfgOptionBooleanDefault("cspResponseHeader", FALSE))
     return;
 written = TRUE;
 generateCspResponseHeader(stdout);
 }