Fatal error: exception Env.Error "Unbound struct 'S'"

[zhendongsu]

$ ccomp -version
The CompCert C verified compiler, version 2.7.1
$
$ ccomp small.c
small.c:1: Error: incomplete type struct S.
Fatal error: exception Env.Error "Unbound struct 'S'"
Raised at file "cparser/Env.ml", line 173, characters 9-48
Called from file "cparser/Cutil.ml", line 147, characters 15-36
Called from file "cparser/Cutil.ml", line 408, characters 28-54
Called from file "cparser/Ceval.ml", line 244, characters 18-32
Called from file "cparser/Ceval.ml", line 267, characters 41-53
Called from file "cparser/Elab.ml", line 625, characters 18-45
Called from file "cparser/Elab.ml", line 727, characters 6-49
Called from file "cparser/Elab.ml", line 93, characters 24-32
Called from file "cparser/Elab.ml", line 734, characters 3-35
Called from file "cparser/Elab.ml", line 2171, characters 8-52
Called from file "cparser/Elab.ml", line 2186, characters 26-54
Called from file "cparser/Elab.ml", line 2441, characters 9-63
Called from file "cparser/Parse.ml", line 76, characters 15-59
Called from file "driver/Frontend.ml", line 58, characters 10-59
Called from file "driver/Driver.ml", line 72, characters 12-41
Called from file "driver/Driveraux.ml", line 140, characters 34-40
Called from file "driver/Driver.ml", line 548, characters 22-70
$
$ cat small.c
int a[alignof(struct S)];
$

[zhendongsu]

Formatting has messed up the code, which should have been:

int a[__alignof__(struct S)];

[bschommer]

Fixed by commit 98427e8.

[zhendongsu]

Thanks for the fix, @bschommer!

BTW, are you able to reproduce #113? I see that it's still open after a few days.

[m-schmidt]

Yes we can reproduce it. Fix in in the works.

[zhendongsu]

Great; thanks.

[xavierleroy]

Please don't rush. I'd like to look at #113 myself but I am currently busy with the death of a close relative.

[m-schmidt]

Regarding this problem: I had a look into IRC.ml. Is it legal to call "IRC.add_pref g a b" with a and b being the same variable?

Michael Schmidt

On 19 Aug 2016, at 19:38, Xavier Leroy notifications@github.com wrote:

Please don't rush. I'd like to look at #113 myself but I am currently busy with the death of a close relative.

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub, or mute the thread.

[xavierleroy]

We really need to move this discussion out of this report. Commit 954b01e plausibly fixes the issue with #113.

Concerning IRC.add_pref a b, I don't think anything goes wrong if a = b. Eventually, IRC.coalesce will drop this move. But clearly this is wasteful in memory (more moves are recorded) and perhaps in speed of register allocation, so we should shortcut this case.