Correction to what I just wrote:
On 25.01.21 11:02, Tobias Burnus wrote:
Executing on $remote_host: rm -f
gcm.cache/$srddir/g++.dg/modules/concept-4.H.gcm (timeout = 360)
→ This deletes in gcm.cache/$srddir
Executing on host: cp -pf /$srcdir/g++.dg/modules/concept-4.H concept-4.H
Executing on host: powerpc64le-none-linux-gnu-g++ $srddir/concept-4.H
... -fmodule-header ...
This should be `pwd` not $srcdir in the g++ call; otherwise, the result
wouldn't be:
→ This creates gcm.cache/`pwd`/concept-5.H.gcm
(`pwd` is our test directory)
Tobias
PS: Full text as it did not seem to show up at gcc-patches@; I
originally had it in BCC instead of CC. (For which I blame the new
mail-client interface.)
On 25.01.21 11:02, Tobias Burnus wrote:
Hi Nathan,
looking at testsuite issues, I see many g++ module fails.
Maybe we can fix some.
* * *
As mentioned on IRC:
09:09 < Tobias__> nathan: I get 'ERROR: can't read "identlink": no
such variable'
'identlink' appears in GCC only once:
09:09 < Tobias__> nathan:
gcc/testsuite/g++.dg/modules/modules.exp: unresolved "$identlink"
09:11 < Tobias__> nathan: Missing space? s/$identlink/$ident link/ ?
* * *
FAIL: g++.dg/modules/bad-mapper-3.C -std=c++17 (test for errors, line )
FAIL: g++.dg/modules/bad-mapper-3.C -std=c++17 (test for excess errors)
Testsuite expects:
// { dg-additional-options "-fmodules-ts
-fmodule-mapper=localhost:172477262" }
// { dg-error {failed connecting mapper 'localhost:172477262'} "" {
target *-*-* } 0 }
but I get:
Excess errors:
.../bad-mapper-3.C: error: failed Name or service not known mapper
'localhost:172477262'
The test expects that Cody::OpenInet6 runs into:
errstr = "connecting";
but it seems to fail already for:
if (int err = getaddrinfo (name, "0", &hints, &addrs))
{
errstr = gai_strerror (err);
The error seems to be:
EAI_NONAME
The node or service is not known; or both node and
service are NULL;
or AI_NUMERICSERV was specified in hints.ai_flags and
service was
not a numeric port-number string.
* * *
There is some issue related to out-of-build-tree testing
and where modules are written:
Executing on $remote_host: rm -f
gcm.cache/$srddir/g++.dg/modules/concept-4.H.gcm (timeout = 360)
→ This deletes in gcm.cache/$srddir
Executing on host: cp -pf /$srcdir/g++.dg/modules/concept-4.H concept-4.H
Executing on host: powerpc64le-none-linux-gnu-g++ $srddir/concept-4.H
... -fmodule-header ...
→ This creates gcm.cache/`pwd`/concept-5.H.gcm
(`pwd` is our test directory)
Executing on $remote_host: rm -f
gcm.cache/$srcdir/g++.dg/modules/concept-4.H.gcm
→ Again $srcdir
This obviously fails unless $srcdir == `pwd`.
I do see testsuite/g++.dg/modules/modules.exp:
setenv CXX_MODULE_PATH "$srcdir/$subdir"
but "CXX_MODULE_PATH" only appears in that file.
(hence: delete or hook it up)
Ideas?
Tobias
-----------------
Mentor Graphics (Deutschland) GmbH, Arnulfstraße 201, 80634 München / Germany
Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Alexander
Walter
