Context : It is still debated whether or not the Galactic chemical evolution of sulphur in the halo followed the constant or flat trend with [ \mathrm { Fe } / \mathrm { H } ] , ascribed to the result of explosive nucleosynthesis in type II SNe .
It has been suggested that the disagreement between different investigations of sulphur abundances in halo stars might be due to problems with the diagnostics used , that a new production source of sulphur might be needed in the early Universe , like hypernovae , or that the deposition of supernova ejecta into the interstellar medium is time-delayed .

Aims : The aim of this study is to try to clarify this situation by measuring the sulphur abundance in a sample of halo giants using two diagnostics ; the S i triplet around 1045 nm and the [ S i ] line at 1082 nm .
The latter of the two is not believed to be sensitive to non-LTE effects .
We can thereby minimize the uncertainties in the diagnostic used and estimate the usefulness of the triplet in sulphur determination in halo K giants .
We will also be able to compare our sulphur abundance differences from the two diagnostics with the expected non-LTE effects in the 1045 nm triplet previously calculated by others .

Methods : High-resolution near-infrared spectra of ten K giants were recorded using the spectrometer CRIRES mounted on VLT .
Two standard settings were used ; one covering the S i triplet and one covering the [ S i ] line .
The sulphur abundances were determined individually with equivalent widths and synthetic spectra for the two diagnostics using tailored 1D model atmospheres and relying on non-LTE corrections from the litterature .
Effects of convective inhomogeneities in the stellar atmospheres are investigated .

Results : The sulphur abundances derived from both the 1082 nm [ S i ] line and the non-LTE corrected 1045 nm triplet favor a flat trend for the evolution of sulphur .
No ‘ high ’ values of [ \mathrm { S } / \mathrm { Fe } ] shown in some previous studies are seen in our sample .

Conclusions : We corroborate the flat trend in the [ \mathrm { S } / \mathrm { Fe } ] vs . [ \mathrm { Fe } / \mathrm { H } ] plot for halo stars found in other works and can not find a scatter nor a rise in [ \mathrm { S } / \mathrm { Fe } ] obtained in some other previous studies .
We find the sulphur abundances deduced from the non-LTE corrected triplet somewhat lower than the abundances from the [ S i ] line , possibly indicating too large non-LTE corrections .
Considering 3D modeling , however , they might instead be too small .
Further we show that the [ S i ] line is possible to use as a sulphur diagnostic down to [ \mathrm { Fe } / \mathrm { H } ] \sim - 2.3 in giants .