This is the mail archive of the binutils@sources.redhat.com mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[PATCH RFA] get the MIPS ECOFF monkey off my back.


So, over the past N months i'd occasionally run the testsuite for the
mips-ecoff target.

I don't have a MIPS ECOFF system, I was just doing it to see if any
changes I was making (mostly to the embedded-pic code) broke anything.

Alas, the MIPS ECOFF testsuite, especially for embedded-pic code, is
broken, and is getting more broken all the time.  It's no longer
useful for me to try my tests there, but I do want to pass of whatever
knowledge I have about the failing tests.


The patch below documents the current state of brokenness in the MIPS
ECOFF tests.  It's there in the hope that somebody will look at them
one day, and that the observations/bug-tracking i've done will come in
handy.

What say you?


FWIW, at some point "we" might want to consider completely splitting,
either into different .exp files, different directories, or at least
different big 'if' sections of mips.exp.  In my experience, lately, if
a given test used on multiple object formats is changed or added, odds
are that in fact it doesn't work on ECOFF.  I can only imagine what
the current situation is like on an a.out target, and you know what,
I'm not going to try to find out.  8-)



cgd
===================================================================
[gas/testsuite/ChangeLog]
2002-02-08  Chris Demetriou  <cgd@broadcom.com>

	* gas/mips/mips.exp: Document (but do not XFAIL) currently-failing
	MIPS ECOFF tests, providing what information is known about
	the origins of the failures.

Index: gas/mips/mips.exp
===================================================================
RCS file: /cvs/src/src/gas/testsuite/gas/mips/mips.exp,v
retrieving revision 1.27
diff -u -r1.27 mips.exp
--- mips.exp	2002/01/02 18:59:30	1.27
+++ mips.exp	2002/02/09 02:52:46
@@ -2,6 +2,12 @@
 # Some generic MIPS tests
 #
 
+# "LOSE" marks information about tests which fail at a particular point
+# in time, but which are not XFAILed.  Either they used to pass
+# and indicate either regressions or the need to tweak the tests to keep
+# up the with code, or they are new tests and it is unknown whether or not
+# they should pass as-is for the given object formats.
+
 proc run_list_test { name opts } {
     global srcdir subdir
     set testname "MIPS $name"
@@ -40,11 +46,17 @@
     run_dump_test "and"
     run_dump_test "break20"
     run_dump_test "trap20"
+
+    # LOSE: As of 2002-02-08, "beq" through "bltu" fail for target mips-ecoff.
+    # See http://sources.redhat.com/ml/binutils/2001-10/msg00418.html for
+    # more information.  Not sure if fixes there are correct; should
+    # branches to external labels be allowed for ECOFF?
     run_dump_test "beq"
     run_dump_test "bge"
     run_dump_test "bgeu"
     run_dump_test "blt"
     run_dump_test "bltu"
+
     if !$ilocks { run_dump_test "div" } else { run_dump_test "div-ilocks" }
     run_dump_test "dli"
     if $elf {
@@ -54,10 +66,15 @@
     }
     if $elf { run_dump_test "jal-svr4pic" }
     if $elf { run_dump_test "jal-xgot" }
+    # LOSE: As of 2002-02-08, the jal-empic test fails.  It appears that it
+    # broke between 2000-03-11 00:00UTC and 2000-03-12 00:00 UTC.
     if $ecoff { run_dump_test "jal-empic" }
     if !$aout { run_dump_test "la" }
     if $elf { run_dump_test "la-svr4pic" }
     if $elf { run_dump_test "la-xgot" }
+    # LOSE: As of 2002-02-08, the la-empic test fails.  Not sure when it
+    # first cropped up, but may be related to addition of "la" -> "addiu"
+    # pattern in MIPS opcode table long ago.
     if $ecoff { run_dump_test "la-empic" }
     if !$aout { run_dump_test "lb" }
     if $elf { run_dump_test "lb-svr4pic" }
@@ -86,6 +103,8 @@
     if !$aout { run_dump_test "lifloat" }
     if $elf { run_dump_test "lif-svr4pic" }
     if $elf { run_dump_test "lif-xgot" }
+    # LOSE: As of 2002-02-08, the lif-empic test fails.  It appears that it
+    # broke between 2000-03-11 00:00UTC and 2000-03-12 00:00 UTC.
     if $ecoff { run_dump_test "lif-empic" }
     run_dump_test "mips4"
     if !$ilocks { run_dump_test "mul" } else { run_dump_test "mul-ilocks" }
@@ -126,6 +145,9 @@
 
     run_list_test "illegal" ""
 
+    # LOSE: As of 2002-02-08, the next 4 tests fail for target mips-ecoff.
+    # It's unknown whether they _should_ pass as-is, or whether different
+    # variants are needed for ELF and ECOFF.
     run_dump_test "mips-gp32-fp32"
     run_dump_test "mips-gp32-fp64"
     run_dump_test "mips-gp64-fp32"


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]