l_push_char allocates memory which is freed with free_line. However, 
currently, the memory is not always freed when calling generate_error.

If one aborts, that's fine. However, generate_error can also set the 
iostat variable. Thus, one should ensure that the memory is always freed.

I wouldn't mind if someone, who knows libgfortran/io better than I, 
could confirm that the patch is okay. I don't want too free memory which 
is later used.

Build and regtested on x86-64-linux.
OK for the trunk?

Tobias


PS: Test case see PR. (I don't think it makes sense to include it in the 
test suite.)

PPS: I see the following failures when regtesting; the LTO and the 
realloc one I know, but I haven't see the reassoc_4.f before. It cannot 
be due to my patch and when I run it manually, it also works:

FAIL: gfortran.dg/lto/pr45586 f_lto_pr45586_0.o-f_lto_pr45586_0.o link, 
-O0 -flto  (internal compiler error)
FAIL: gfortran.dg/realloc_on_assign_5.f03  -O0  execution test
FAIL: gfortran.dg/reassoc_4.f  -O   scan-tree-dump-times reassoc1 "[0-9] 
\\* " 22