4D-API_cs2cs-style.gie fails on Debian unstable (with GCC 8) #1084
Milestone
Comments
rouault
added a commit
to rouault/PROJ
that referenced
this issue
Aug 10, 2018
torad_coord() of gie.c has this sequence:
```
if( cond )
axis = l->param + strlen ("axis=");
n = strlen (axis);
```
When the if branch is evaluated, n is always zero
even if on inspection axis is non empty
The reason is that the struct ARG_list which is the
type of l use a variable-length array for the param member
struct ARG_list {
paralist *next;
char used;
char param[1];
};
But this is not a proper way of declaring it, and
gcc 8 has apparently optimizations to detect that l->param + 5
points out of the array, hence it optimizes strlen() to 0.
According to https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html,
the proper way of declaring such arrays is to use [0]
rouault
added a commit
that referenced
this issue
Aug 11, 2018
Fix wrong behaviour of torad_coord() with gcc 8.1 -O2 (fixes #1084)
|
@sebastic. Fixed. I suspect a gcc optimization bug (reported as https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86914 ) |
|
Thanks for the patch and in-depth analysis. I've included the changes from 143b4d3 as a patch in the 5.1.0 Debian package, but the test still fails in the same way: Full buildlog: proj_5.1.0-2_amd64.txt |
|
@sebastic Weird. Can you just try to build master and make check (after installing egm96_15.gtx in nad/) with the same compiler ? |
|
|
|
OK, I can reproduce within a Debian unstable VM with gcc 8.2. Regarding the first failure initially reported (4D-API_cs2cs-style.gie(114)), this is a difference between gcc 8.1 and 8.2, and the later only generates the correct code if using the [] syntax (I've updated https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86914 with it). Investigating for the new issue: 4D-API_cs2cs-style.gie(256) |
|
Confirmed fixed with the changes from edfa298, thanks! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4D-API_cs2cs-style.giestarted to fail on Debian unstable recently. Quite possibly related to GCC 8 becoming the default compiler.The text was updated successfully, but these errors were encountered: