lkml.org 
[lkml]   [2020]   [May]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: I disabled more compiler warnings..
On Mon, May 11, 2020 at 11:03 AM Arnd Bergmann <arnd@arndb.de> wrote:
>
> - There is a good chance that some trivial difference between
> mainline and linux-next, or between linux-next and my test
> branch (with a couple hundred fixup patches) is causing you
> to see many warnings that I don't get. I'll try to reproduce and
> bisect what I find.

Update: I found a bug in my own patches, one of them tried to disable
the warning when CONFIG_GCOV_PROFILE_ALL is turned on, as that
caused some false positives. Unfortunately I ended up turning it off
unconditionally by testing for CONFIG_ARCH_HAS_GCOV_PROFILE_ALL
instead.

I do see 36 such errors in my tree now with gcc-10 x86 allmodconfig,
and there are probably a couple more in your tree that I had already
fixed locally, e.g. when clang found the same thing. See full (abbreviated)
list below.

Looking at some of those warnings, I find a roughly 50:50 split between
code that gcc should have warned about all along when there is a
path in which a variable is actually used without initialization (which
may be impossible for reasons the compiler cannot know), and code
that looks correct to me, but I can see exactly how gcc gets confused,
especially now that it warns about the first class.
It should be possible to avoid these 36 warnings while making the
code more readable in most cases, or adding fake initializations
for false positives without a better workaround.

For -O3, there are many more warnings (225 total in my tree), with
additional valid bugs but also a higher ratio of false positives
in the samples I looked at. gcc-9 -O3 had additional false positives and
also a few extra somewhat sensible warnings, 273 warnings in total,
though these had always been disabled already.

For the default -O2, gcc-9 prints only the three warnings from
kernel/trace/ftrace.c, all of which are false positives that the compiler
should have not warned about IMO.

Arnd

8<---
/* x86 allmodconfig gcc-10 -O2 warnings */
arch/x86/kvm/../../../virt/kvm/kvm_main.c:2441:42: error:
'nr_pages_avail' may be used uninitialized
drivers/video/fbdev/i740fb.c:332:9: error: 'wm' may be used uninitialized
lib/zstd/decompress.c:303:6: error: 'fParams.windowSize' may be used
uninitialized
lib/zstd/decompress.c:342:8: error: 'fParams.frameContentSize' may be
used uninitialized
kernel/trace/ftrace.c:7156:8: error: 'seq_ops' may be used uninitialized
kernel/trace/ftrace.c:7234:8: error: 'filtered_pids' may be used uninitialized
kernel/trace/ftrace.c:7251:22: error: 'other_pids' may be used uninitialized
drivers/i3c/device.c:227:41: error: 'devinfo.dcr' may be used uninitialized
include/linux/i3c/device.h:80:49: error: 'devinfo.pid' may be used uninitialized
fs/ext4/ext4_extents.h:226:8: error: 'zero_ex1.ee_start_lo' may be
used uninitialized
fs/ext4/ext4_extents.h:227:12: error: 'zero_ex1.ee_start_hi' may be
used uninitialized
fs/ext4/extents.c:3083:9: error: 'zero_ex1.ee_block' may be used uninitialized
drivers/input/touchscreen/hideep.c:373:14: error: 'unmask_code' may be
used uninitialized
drivers/media/i2c/tvp7002.c:691:11: error: 'val' may be used uninitialized
include/linux/spinlock.h:289:3: error: 'flags' may be used uninitialized
drivers/gpu/drm/amd/amdgpu/df_v3_6.c:571:2: error: 'hi_base_addr' may
be used uninitialized
drivers/gpu/drm/amd/amdgpu/df_v3_6.c:571:2: error: 'lo_base_addr' may
be used uninitialized
drivers/gpu/drm/i915/gt/selftest_lrc.c:2959:28: error: 'rq' may be
used uninitialized
drivers/gpu/drm/i915/gt/selftest_workarounds.c:492:17: error: 'rsvd'
may be used uninitialized
drivers/firmware/arm_scmi/driver.c:502:12: error: 't' may be used uninitialized
drivers/usb/serial/cp210x.c:948:8: error: 'bits' may be used uninitialized
drivers/usb/serial/cp210x.c:1194:8: error: 'bits' may be used uninitialized
drivers/platform/x86/acer-wmi.c:1440:9: error: 'value' may be used uninitialized
drivers/net/wireless/rsi/rsi_91x_sdio.c:1340:7: error: 'data' may be
used uninitialized
drivers/net/wireless/rsi/rsi_91x_sdio.c:238:12: error: 'resp' may be
used uninitialized
drivers/gpu/drm/nouveau/nvkm/subdev/i2c/auxg94.c:160:45: error: 'stat'
may be used uninitialized
drivers/gpu/drm/nouveau/nvkm/subdev/i2c/auxgm200.c:160:45: error:
'stat' may be used uninitialized
drivers/ata/ahci_imx.c:369:30: error: 'dac_ctl_reg' may be used uninitialized
drivers/ata/ahci_imx.c:377:33: error: 'rtune_ctl_reg' may be used uninitialized
drivers/ata/ahci_imx.c:375:42: error: 'mpll_test_reg' may be used uninitialized
include/linux/printk.h:360:2: error: 'abs_ppfid' may be used uninitialized
drivers/media/dvb-frontends/drxk_hard.c:1026:5: error: 'wait_cmd' may
be used uninitialized
drivers/iio/imu/bmi160/bmi160_core.c:577:47: error:
'int_out_ctrl_shift' may be used uninitialized
include/linux/dev_printk.h:104:2: error: 'pin_name' may be used uninitialized
drivers/iio/imu/bmi160/bmi160_core.c:606:8: error: 'int_map_mask' may
be used uninitialized
drivers/iio/imu/bmi160/bmi160_core.c:599:8: error: 'int_latch_mask'
may be used uninitialized

\
 
 \ /
  Last update: 2020-05-11 13:12    [W:0.086 / U:0.048 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site