unsigned int vs long in RHEL libconfig.

xava-charlesj commented

2019-03-19 13:00:33 +01:00

(Migrated from github.com)

Not sure if it's just RHEL 6, but I had issues with all the unsigned int in the config struct, the config_lookup_int libarary has long as param 3 - so it was trashing 4 bytes of the next config option in the master config structure each time it got a value. Fixed by changing all appropriate unsigned ints to longs in the .h. (refresh_token_expiration was being trashed by a later grab of session_expiration, since it sits just above it in the .h struct). Code didn't crash but it appeared as a bug where refresh_token_expiration was always seemingly zero despite config being set. Walking through the code highlighted it was being set to 0 by:

config_lookup_int(&cfg, "session_expiration", &config->session_expiration);

Long params as per libconfig.h

extern LIBCONFIG_API int config_lookup_int(const config_t *config,
const char *path, long *value);

sizeof unsigned int: 4
sizeof long: 8.

Not sure if it's just RHEL 6, but I had issues with all the unsigned int in the config struct, the config_lookup_int libarary has long as param 3 - so it was trashing 4 bytes of the next config option in the master config structure each time it got a value. Fixed by changing all appropriate unsigned ints to longs in the .h. (refresh_token_expiration was being trashed by a later grab of session_expiration, since it sits just above it in the .h struct). Code didn't crash but it appeared as a bug where refresh_token_expiration was always seemingly zero despite config being set. Walking through the code highlighted it was being set to 0 by: config_lookup_int(&cfg, "session_expiration", &config->session_expiration); Long params as per libconfig.h extern LIBCONFIG_API int config_lookup_int(const config_t *config, const char *path, long *value); sizeof unsigned int: 4 sizeof long: 8.

babelouest commented

2019-03-19 15:29:24 +01:00

(Migrated from github.com)

Glewlwyd is built with libconfig 1.4, where, according to the documentation expects an int in the function config_lookup_int: https://www.hyperrealm.com/libconfig/libconfig_manual.html#The-C-API

So I prefer to stay in the current API in that case.

Maybe you should install a more recent version of libconfig?

Glewlwyd is built with libconfig 1.4, where, according to the documentation expects an int in the function `config_lookup_int`: https://www.hyperrealm.com/libconfig/libconfig_manual.html#The-C-API So I prefer to stay in the current API in that case. Maybe you should install a more recent version of libconfig?

xava-charlesj commented

2019-03-19 22:00:58 +01:00

(Migrated from github.com)

You're absolutely right - apologies, the perils of legacy environments. Could it be plausible to add a Make / compiler warning to suit if libconfig < 1.4? Given that the effects are so subtle and may well go unnoticed for some.

babelouest commented

2019-03-20 00:18:39 +01:00

(Migrated from github.com)

I will add a check in the CMakelists.txt file and add a warning in the documentation, thanks for the report.

I will add a check in the `CMakelists.txt` file and add a warning in the documentation, thanks for the report.

babelouest commented

2019-03-20 01:37:19 +01:00

(Migrated from github.com)

@xava-charlesj , this should be fixed in the branch 1.4 now, thanks

👍 1

xava-charlesj commented

2019-03-20 01:39:08 +01:00

(Migrated from github.com)

Great, thank you - appreciate that it's a little trivial.

FYI, am currently running 1.4.10 alongside libconfig 1.7.2 and all is well so far.

Great, thank you - appreciate that it's a little trivial. FYI, am currently running 1.4.10 alongside libconfig 1.7.2 and all is well so far.

babelouest commented

2019-03-20 01:46:32 +01:00

(Migrated from github.com)

Awesome, I'm closing the issue then

Rows
Columns

unsigned int vs long in RHEL libconfig. #55