Wrong sem_id in semtest

[eternalNight]

Description

In the unit test semtest, the sem_id doesn't meet the constraint in the test spec file.

Commit

7e99100 unittest: Adapt tests on sfs to the current sfsimg.

Reproduce Steps

cd ucore
make ARCH=i386 menuconfig
UCORE_TEST=/testbin/semtest make kernel
UCORE_TEST=/testbin/semtest make sfsimg
make swapimg
./uCore_test -sf src/user-ucore/testspecs/semtest.testspec

Current Result

The unit test fail with

!! error: missing 'sem_id = 0x07......'

In serial.log, the sem_id is

sem_id = 0x000000011fe625e0

Expected Result

The unit test doesn't fail with the line mentioned above.
Either provide a fix for the kernel or update the test spec with reasonable analysis.

[chyh1990]

I change the sem_id base to a large integer.
Originally, sem_id started from 0 is a problem

[eternalNight]

If so, this may not be an issue on arm as it's sem_id is the 64-bit representation of the required in pattern "0x07......". But the sem_id on i386 still looks strange to me.

BTW, what's the problem caused by sem_id starting from 0?

[chyh1990]

#define VALID_SEMID(sem_id)
((uintptr_t)(sem_id) < (uintptr_t)(sem_id) + PBASE)

#define semid2sem(sem_id)
((semaphore_t *)((uintptr_t)(sem_id) + PBASE))

#define sem2semid(sem)
((sem_t)((uintptr_t)(sem) - PBASE))

/* This must NOT be 0 */
#define PBASE (KERNBASE+0x100)

on some platforms, KERNBASE may be 0, so i add 0x100 to it

On Sun, Sep 9, 2012 at 12:21 PM, Mao Junjie notifications@github.comwrote:

If so, this may not be an issue on arm as it's sem_id is the 64-bit
representation of the required in pattern "0x07......". But the sem_id on
i386 still looks strange to me.

BTW, what's the problem caused by sem_id starting from 0?

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/5#issuecomment-8400567.

Chen Yuheng
Computer Science
Tsinghua University