[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Testing on HEAD, summary
From: |
Hans-Peter Nilsson |
Subject: |
Testing on HEAD, summary |
Date: |
Sun, 17 Feb 2008 08:45:23 +0100 |
Executive summary: native, sim and board tests say it's fine; at
least no worse than dejagnu-1.4.4; somewhat better. Ship it! ;)
A "git show" says the my latest update is, and verified with
<http://git.savannah.gnu.org/gitweb/?p=dejagnu.git;a=shortlog;h=master>
to be the official latest commit (how do I check this using git?):
commit 94dbf2fc904ce3446aa513e41edac6a3edaf316d
Author: Ben Elliston <address@hidden>
Date: Mon Feb 11 21:48:59 2008 +1100
With this, I've run on x86_64-unknown-linux-gnu Debian etch, and
compared to their dejagnu package (which includes "lots" of
current HEAD, so I guess no changes were really expected):
- native bootstrap+check testing binutils and gcc trunk
r132182 (same revision for tests below).
- simulator setups (as stated or implicitly target X-elf +
baseboard X-sim; included in dejagnu) hosted on the system
above: v850, sh, mmix-knuth-mmixware+mmix, mips, fr30, arm,
mcore and mipsisa64-elf+mips-sim-idt64 testing binutils, gcc,
newlib and as applicable, sim.
- simulator setups for cris-elf and crisv32-elf with
the originally submitted cris-sim.exp compared to the
cris-sim.exp in HEAD, the rest as above.
- simulator multilib setups for cris-elf with the original
cris-sim.exp compared to the cris-sim.exp in HEAD, i.e.
'RUNTESTFLAGS=--target_board=cris-sim\{,-march=v10,-march=v32\}'.
The rest as above.
- simulator setups for cris-axis-linux-gnu and
crisv32-axis-linux-gnu with the original cris-sim.exp compared
to the cris-sim.exp in HEAD. The rest as above. For crisv32-*
one less libstdc++ test timed out doubtlessly due to differences
in host load.
- sh64-elf with a sh64-sim.exp that Joern Rennecke posted, IIRC.
The rest as above.
For all the above, the only change, was that dejagnu HEAD had an
extra "WARNING: Couldn't find the global config file." for all
testsuites, above the the existing "Test Run By hp on Fri Feb 8
12:36:48 2008" or in some testsuites above the "WARNING:
Couldn't find tool init file".
On a x86_64-unknown-linux-gnu Fedora 8 host, testing cris-elf
with the original cris-sim.exp compared to the cris-sim.exp in
HEAD, against src/sim, binutils and gcc r132266. As posted
before, there are some differences compared to the F 8 dejagnu
package (which is closer to 1.4.4 than the Debian package),
though in all improvements.
- Same host, testing src/sim for (X-elf+X-sim) arm, frv, h8300,
iq2000, m32r, m68hc11, mcore, mips, sh, v850; no changes.
On a x86_64-unknown-linux-gnu Fedora core 4 host targetting
crisv32-axis-linux-gnu with cris-sim and ditto but a
"handcrafted" baseboard file (using telnet and ftp; implementing
telnet_exec "manually") for a crisv32-axis-linux-gnu board
shows, for sim, binutils and gcc r132252, a few more test-cases
passing due to the (as posted) more heavily pruned warnings,
e.g. "fesetenv is not implemented and will always fail" and of
course the marking non-linking run-tests as unresolved. No
other differences attributable to dejagnu.
Those are the variations covered by my own "daily" use with some
additions.
brgds, H-P
- Testing on HEAD, summary,
Hans-Peter Nilsson <=