http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45841
Hans-Peter Nilsson <hp at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Component|libstdc++ |testsuite
--- Comment #15 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2010-10-04
23:51:54 UTC ---
(In reply to comment #9)
> ... something is wrong
> with lseek().
It certainly is! There are at least three issues here.
1. The simulator has a bug: on a 64-bit host, the offset parameter to the lseek
call is in-effect zero-extended from 32-bit-long to 64-bit-long. This causes
any change to the sequence of lseeks to show random regressions.
2. There is a test-suite hook,
libstdc++-v3/testsuite/lib/libstdc++.exp:check_v3_target_fileio that should
catch this.
3. Fixing #1 certainly shows improvement; 14 additional tests pass, but also 2
regressions, so there's still another issue...
Due to (#1 and) #2, I'm recategorizing this PR as belonging to the testsuite
and I'm assigning it to myself, pending a patch for #2 and investigating #3.
Sorry for the noise, feel free to remove yourself from CC if this made you lose
interest. :)
For the record, an strace excerpt of the simulator before and after r164529
shows the issue better than the actual simulator trace I had in mind. (Though
it also shows that the failing ltrace call is an addition, and I guess you
would want to eliminate it, if possible. Oops, four issues! Maybe that was
what the attached patch was about...)
r164528:
...
open("seekoff-2io.tst", O_RDWR) = 4
fstat(4, {st_mode=S_IFREG|0644, st_size=116, ...}) = 0
lseek(4, 0, SEEK_CUR) = 0
lseek(4, 2, SEEK_SET) = 2
fstat(4, {st_mode=S_IFREG|0644, st_size=116, ...}) = 0
lseek(4, 0, SEEK_CUR) = 2
read(4, " ", 1) = 1
read(4, "9", 1) = 1
lseek(4, 3, SEEK_SET) = 3
write(4, "\n", 1) = 1
lseek(4, 4, SEEK_SET) = 4
read(4, "9", 1) = 1
lseek(4, 1, SEEK_CUR) = 6
read(4, "1", 1) = 1
read(4, "1", 1) = 1
lseek(4, 4294967295, SEEK_CUR) = 4294967303
write(4, "x", 1) = 1
write(4, "\n", 1) = 1
lseek(4, 0, SEEK_CUR) = 4294967305
read(4, "", 1) = 0
read(4, "", 1) = 0
lseek(4, 0, SEEK_END) = 4294967305
write(4, "\n", 1) = 1
write(4, "because because because. . .", 28) = 28
lseek(4, 4294967295, SEEK_END) = 8589934629
read(4, "", 1) = 0
lseek(4, 4294967295, SEEK_CUR) = 12884901924
read(4, "", 1) = 0
read(4, "", 1) = 0
read(4, "", 1) = 0
close(4) = 0
...
r164529:
...
open("seekoff-2io.tst", O_RDWR) = 4
fstat(4, {st_mode=S_IFREG|0644, st_size=116, ...}) = 0
lseek(4, 0, SEEK_CUR) = 0
lseek(4, 2, SEEK_SET) = 2
fstat(4, {st_mode=S_IFREG|0644, st_size=116, ...}) = 0
lseek(4, 0, SEEK_CUR) = 2
read(4, " ", 1) = 1
read(4, "9", 1) = 1
lseek(4, 3, SEEK_SET) = 3
write(4, "\n", 1) = 1
lseek(4, 4, SEEK_SET) = 4
read(4, "9", 1) = 1
lseek(4, 1, SEEK_CUR) = 6
read(4, "1", 1) = 1
read(4, "1", 1) = 1
lseek(4, 0, SEEK_CUR) = 8
lseek(4, 4294967295, SEEK_CUR) = 4294967303
write(4, "x", 1) = 1
write(4, "\n", 1) = 1
lseek(4, 0, SEEK_CUR) = 4294967305
read(4, "", 1) = 0
read(4, "", 1) = 0
write(2, "assertion \"", 11assertion ") = 11
write(2, "c1 == c3", 8c1 == c3) = 8
write(2, "\" failed: file \"", 16" failed: file ") = 16
...
